hw/nvme: flexible data placement emulation

Add emulation of TP4146 ("Flexible Data Placement").

Reviewed-by: Keith Busch <kbusch@kernel.org>
Signed-off-by: Jesper Devantier <j.devantier@samsung.com>
Signed-off-by: Klaus Jensen <k.jensen@samsung.com>
This commit is contained in:
Jesper Devantier 2023-02-20 12:59:26 +01:00 committed by Klaus Jensen
parent e181d3da39
commit 73064edfb8
6 changed files with 1179 additions and 15 deletions

View file

@ -238,6 +238,8 @@ static const bool nvme_feature_support[NVME_FID_MAX] = {
[NVME_TIMESTAMP] = true,
[NVME_HOST_BEHAVIOR_SUPPORT] = true,
[NVME_COMMAND_SET_PROFILE] = true,
[NVME_FDP_MODE] = true,
[NVME_FDP_EVENTS] = true,
};
static const uint32_t nvme_feature_cap[NVME_FID_MAX] = {
@ -249,6 +251,8 @@ static const uint32_t nvme_feature_cap[NVME_FID_MAX] = {
[NVME_TIMESTAMP] = NVME_FEAT_CAP_CHANGE,
[NVME_HOST_BEHAVIOR_SUPPORT] = NVME_FEAT_CAP_CHANGE,
[NVME_COMMAND_SET_PROFILE] = NVME_FEAT_CAP_CHANGE,
[NVME_FDP_MODE] = NVME_FEAT_CAP_CHANGE,
[NVME_FDP_EVENTS] = NVME_FEAT_CAP_CHANGE | NVME_FEAT_CAP_NS,
};
static const uint32_t nvme_cse_acs[256] = {
@ -281,6 +285,8 @@ static const uint32_t nvme_cse_iocs_nvm[256] = {
[NVME_CMD_VERIFY] = NVME_CMD_EFF_CSUPP,
[NVME_CMD_COPY] = NVME_CMD_EFF_CSUPP | NVME_CMD_EFF_LBCC,
[NVME_CMD_COMPARE] = NVME_CMD_EFF_CSUPP,
[NVME_CMD_IO_MGMT_RECV] = NVME_CMD_EFF_CSUPP,
[NVME_CMD_IO_MGMT_SEND] = NVME_CMD_EFF_CSUPP | NVME_CMD_EFF_LBCC,
};
static const uint32_t nvme_cse_iocs_zoned[256] = {
@ -299,12 +305,66 @@ static const uint32_t nvme_cse_iocs_zoned[256] = {
static void nvme_process_sq(void *opaque);
static void nvme_ctrl_reset(NvmeCtrl *n, NvmeResetType rst);
static inline uint64_t nvme_get_timestamp(const NvmeCtrl *n);
static uint16_t nvme_sqid(NvmeRequest *req)
{
return le16_to_cpu(req->sq->sqid);
}
static inline uint16_t nvme_make_pid(NvmeNamespace *ns, uint16_t rg,
uint16_t ph)
{
uint16_t rgif = ns->endgrp->fdp.rgif;
if (!rgif) {
return ph;
}
return (rg << (16 - rgif)) | ph;
}
static inline bool nvme_ph_valid(NvmeNamespace *ns, uint16_t ph)
{
return ph < ns->fdp.nphs;
}
static inline bool nvme_rg_valid(NvmeEnduranceGroup *endgrp, uint16_t rg)
{
return rg < endgrp->fdp.nrg;
}
static inline uint16_t nvme_pid2ph(NvmeNamespace *ns, uint16_t pid)
{
uint16_t rgif = ns->endgrp->fdp.rgif;
if (!rgif) {
return pid;
}
return pid & ((1 << (15 - rgif)) - 1);
}
static inline uint16_t nvme_pid2rg(NvmeNamespace *ns, uint16_t pid)
{
uint16_t rgif = ns->endgrp->fdp.rgif;
if (!rgif) {
return 0;
}
return pid >> (16 - rgif);
}
static inline bool nvme_parse_pid(NvmeNamespace *ns, uint16_t pid,
uint16_t *ph, uint16_t *rg)
{
*rg = nvme_pid2rg(ns, pid);
*ph = nvme_pid2ph(ns, pid);
return nvme_ph_valid(ns, *ph) && nvme_rg_valid(ns->endgrp, *rg);
}
static void nvme_assign_zone_state(NvmeNamespace *ns, NvmeZone *zone,
NvmeZoneState state)
{
@ -378,6 +438,69 @@ static uint16_t nvme_aor_check(NvmeNamespace *ns, uint32_t act, uint32_t opn)
return nvme_zns_check_resources(ns, act, opn, 0);
}
static NvmeFdpEvent *nvme_fdp_alloc_event(NvmeCtrl *n, NvmeFdpEventBuffer *ebuf)
{
NvmeFdpEvent *ret = NULL;
bool is_full = ebuf->next == ebuf->start && ebuf->nelems;
ret = &ebuf->events[ebuf->next++];
if (unlikely(ebuf->next == NVME_FDP_MAX_EVENTS)) {
ebuf->next = 0;
}
if (is_full) {
ebuf->start = ebuf->next;
} else {
ebuf->nelems++;
}
memset(ret, 0, sizeof(NvmeFdpEvent));
ret->timestamp = nvme_get_timestamp(n);
return ret;
}
static inline int log_event(NvmeRuHandle *ruh, uint8_t event_type)
{
return (ruh->event_filter >> nvme_fdp_evf_shifts[event_type]) & 0x1;
}
static bool nvme_update_ruh(NvmeCtrl *n, NvmeNamespace *ns, uint16_t pid)
{
NvmeEnduranceGroup *endgrp = ns->endgrp;
NvmeRuHandle *ruh;
NvmeReclaimUnit *ru;
NvmeFdpEvent *e = NULL;
uint16_t ph, rg, ruhid;
if (!nvme_parse_pid(ns, pid, &ph, &rg)) {
return false;
}
ruhid = ns->fdp.phs[ph];
ruh = &endgrp->fdp.ruhs[ruhid];
ru = &ruh->rus[rg];
if (ru->ruamw) {
if (log_event(ruh, FDP_EVT_RU_NOT_FULLY_WRITTEN)) {
e = nvme_fdp_alloc_event(n, &endgrp->fdp.host_events);
e->type = FDP_EVT_RU_NOT_FULLY_WRITTEN;
e->flags = FDPEF_PIV | FDPEF_NSIDV | FDPEF_LV;
e->pid = cpu_to_le16(pid);
e->nsid = cpu_to_le32(ns->params.nsid);
e->rgid = cpu_to_le16(rg);
e->ruhid = cpu_to_le16(ruhid);
}
/* log (eventual) GC overhead of prematurely swapping the RU */
nvme_fdp_stat_inc(&endgrp->fdp.mbmw, nvme_l2b(ns, ru->ruamw));
}
ru->ruamw = ruh->ruamw;
return true;
}
static bool nvme_addr_is_cmb(NvmeCtrl *n, hwaddr addr)
{
hwaddr hi, lo;
@ -3322,6 +3445,41 @@ invalid:
return status | NVME_DNR;
}
static void nvme_do_write_fdp(NvmeCtrl *n, NvmeRequest *req, uint64_t slba,
uint32_t nlb)
{
NvmeNamespace *ns = req->ns;
NvmeRwCmd *rw = (NvmeRwCmd *)&req->cmd;
uint64_t data_size = nvme_l2b(ns, nlb);
uint32_t dw12 = le32_to_cpu(req->cmd.cdw12);
uint8_t dtype = (dw12 >> 20) & 0xf;
uint16_t pid = le16_to_cpu(rw->dspec);
uint16_t ph, rg, ruhid;
NvmeReclaimUnit *ru;
if (dtype != NVME_DIRECTIVE_DATA_PLACEMENT ||
!nvme_parse_pid(ns, pid, &ph, &rg)) {
ph = 0;
rg = 0;
}
ruhid = ns->fdp.phs[ph];
ru = &ns->endgrp->fdp.ruhs[ruhid].rus[rg];
nvme_fdp_stat_inc(&ns->endgrp->fdp.hbmw, data_size);
nvme_fdp_stat_inc(&ns->endgrp->fdp.mbmw, data_size);
while (nlb) {
if (nlb < ru->ruamw) {
ru->ruamw -= nlb;
break;
}
nlb -= ru->ruamw;
nvme_update_ruh(n, ns, pid);
}
}
static uint16_t nvme_do_write(NvmeCtrl *n, NvmeRequest *req, bool append,
bool wrz)
{
@ -3431,6 +3589,8 @@ static uint16_t nvme_do_write(NvmeCtrl *n, NvmeRequest *req, bool append,
if (!(zone->d.za & NVME_ZA_ZRWA_VALID)) {
zone->w_ptr += nlb;
}
} else if (ns->endgrp && ns->endgrp->fdp.enabled) {
nvme_do_write_fdp(n, req, slba, nlb);
}
data_offset = nvme_l2b(ns, slba);
@ -4088,6 +4248,126 @@ static uint16_t nvme_zone_mgmt_recv(NvmeCtrl *n, NvmeRequest *req)
return status;
}
static uint16_t nvme_io_mgmt_recv_ruhs(NvmeCtrl *n, NvmeRequest *req,
size_t len)
{
NvmeNamespace *ns = req->ns;
NvmeEnduranceGroup *endgrp;
NvmeRuhStatus *hdr;
NvmeRuhStatusDescr *ruhsd;
unsigned int nruhsd;
uint16_t rg, ph, *ruhid;
size_t trans_len;
g_autofree uint8_t *buf = NULL;
if (!n->subsys) {
return NVME_INVALID_FIELD | NVME_DNR;
}
if (ns->params.nsid == 0 || ns->params.nsid == 0xffffffff) {
return NVME_INVALID_NSID | NVME_DNR;
}
if (!n->subsys->endgrp.fdp.enabled) {
return NVME_FDP_DISABLED | NVME_DNR;
}
endgrp = ns->endgrp;
nruhsd = ns->fdp.nphs * endgrp->fdp.nrg;
trans_len = sizeof(NvmeRuhStatus) + nruhsd * sizeof(NvmeRuhStatusDescr);
buf = g_malloc(trans_len);
trans_len = MIN(trans_len, len);
hdr = (NvmeRuhStatus *)buf;
ruhsd = (NvmeRuhStatusDescr *)(buf + sizeof(NvmeRuhStatus));
hdr->nruhsd = cpu_to_le16(nruhsd);
ruhid = ns->fdp.phs;
for (ph = 0; ph < ns->fdp.nphs; ph++, ruhid++) {
NvmeRuHandle *ruh = &endgrp->fdp.ruhs[*ruhid];
for (rg = 0; rg < endgrp->fdp.nrg; rg++, ruhsd++) {
uint16_t pid = nvme_make_pid(ns, rg, ph);
ruhsd->pid = cpu_to_le16(pid);
ruhsd->ruhid = *ruhid;
ruhsd->earutr = 0;
ruhsd->ruamw = cpu_to_le64(ruh->rus[rg].ruamw);
}
}
return nvme_c2h(n, buf, trans_len, req);
}
static uint16_t nvme_io_mgmt_recv(NvmeCtrl *n, NvmeRequest *req)
{
NvmeCmd *cmd = &req->cmd;
uint32_t cdw10 = le32_to_cpu(cmd->cdw10);
uint32_t numd = le32_to_cpu(cmd->cdw11);
uint8_t mo = (cdw10 & 0xff);
size_t len = (numd + 1) << 2;
switch (mo) {
case NVME_IOMR_MO_NOP:
return 0;
case NVME_IOMR_MO_RUH_STATUS:
return nvme_io_mgmt_recv_ruhs(n, req, len);
default:
return NVME_INVALID_FIELD | NVME_DNR;
};
}
static uint16_t nvme_io_mgmt_send_ruh_update(NvmeCtrl *n, NvmeRequest *req)
{
NvmeCmd *cmd = &req->cmd;
NvmeNamespace *ns = req->ns;
uint32_t cdw10 = le32_to_cpu(cmd->cdw10);
uint16_t ret = NVME_SUCCESS;
uint32_t npid = (cdw10 >> 1) + 1;
unsigned int i = 0;
g_autofree uint16_t *pids = NULL;
uint32_t maxnpid = n->subsys->endgrp.fdp.nrg * n->subsys->endgrp.fdp.nruh;
if (unlikely(npid >= MIN(NVME_FDP_MAXPIDS, maxnpid))) {
return NVME_INVALID_FIELD | NVME_DNR;
}
pids = g_new(uint16_t, npid);
ret = nvme_h2c(n, pids, npid * sizeof(uint16_t), req);
if (ret) {
return ret;
}
for (; i < npid; i++) {
if (!nvme_update_ruh(n, ns, pids[i])) {
return NVME_INVALID_FIELD | NVME_DNR;
}
}
return ret;
}
static uint16_t nvme_io_mgmt_send(NvmeCtrl *n, NvmeRequest *req)
{
NvmeCmd *cmd = &req->cmd;
uint32_t cdw10 = le32_to_cpu(cmd->cdw10);
uint8_t mo = (cdw10 & 0xff);
switch (mo) {
case NVME_IOMS_MO_NOP:
return 0;
case NVME_IOMS_MO_RUH_UPDATE:
return nvme_io_mgmt_send_ruh_update(n, req);
default:
return NVME_INVALID_FIELD | NVME_DNR;
};
}
static uint16_t nvme_io_cmd(NvmeCtrl *n, NvmeRequest *req)
{
NvmeNamespace *ns;
@ -4164,6 +4444,10 @@ static uint16_t nvme_io_cmd(NvmeCtrl *n, NvmeRequest *req)
return nvme_zone_mgmt_send(n, req);
case NVME_CMD_ZONE_MGMT_RECV:
return nvme_zone_mgmt_recv(n, req);
case NVME_CMD_IO_MGMT_RECV:
return nvme_io_mgmt_recv(n, req);
case NVME_CMD_IO_MGMT_SEND:
return nvme_io_mgmt_send(n, req);
default:
assert(false);
}
@ -4623,6 +4907,207 @@ static uint16_t nvme_cmd_effects(NvmeCtrl *n, uint8_t csi, uint32_t buf_len,
return nvme_c2h(n, ((uint8_t *)&log) + off, trans_len, req);
}
static size_t sizeof_fdp_conf_descr(size_t nruh, size_t vss)
{
size_t entry_siz = sizeof(NvmeFdpDescrHdr) + nruh * sizeof(NvmeRuhDescr)
+ vss;
return ROUND_UP(entry_siz, 8);
}
static uint16_t nvme_fdp_confs(NvmeCtrl *n, uint32_t endgrpid, uint32_t buf_len,
uint64_t off, NvmeRequest *req)
{
uint32_t log_size, trans_len;
g_autofree uint8_t *buf = NULL;
NvmeFdpDescrHdr *hdr;
NvmeRuhDescr *ruhd;
NvmeEnduranceGroup *endgrp;
NvmeFdpConfsHdr *log;
size_t nruh, fdp_descr_size;
int i;
if (endgrpid != 1 || !n->subsys) {
return NVME_INVALID_FIELD | NVME_DNR;
}
endgrp = &n->subsys->endgrp;
if (endgrp->fdp.enabled) {
nruh = endgrp->fdp.nruh;
} else {
nruh = 1;
}
fdp_descr_size = sizeof_fdp_conf_descr(nruh, FDPVSS);
log_size = sizeof(NvmeFdpConfsHdr) + fdp_descr_size;
if (off >= log_size) {
return NVME_INVALID_FIELD | NVME_DNR;
}
trans_len = MIN(log_size - off, buf_len);
buf = g_malloc0(log_size);
log = (NvmeFdpConfsHdr *)buf;
hdr = (NvmeFdpDescrHdr *)(log + 1);
ruhd = (NvmeRuhDescr *)(buf + sizeof(*log) + sizeof(*hdr));
log->num_confs = cpu_to_le16(0);
log->size = cpu_to_le32(log_size);
hdr->descr_size = cpu_to_le16(fdp_descr_size);
if (endgrp->fdp.enabled) {
hdr->fdpa = FIELD_DP8(hdr->fdpa, FDPA, VALID, 1);
hdr->fdpa = FIELD_DP8(hdr->fdpa, FDPA, RGIF, endgrp->fdp.rgif);
hdr->nrg = cpu_to_le16(endgrp->fdp.nrg);
hdr->nruh = cpu_to_le16(endgrp->fdp.nruh);
hdr->maxpids = cpu_to_le16(NVME_FDP_MAXPIDS - 1);
hdr->nnss = cpu_to_le32(NVME_MAX_NAMESPACES);
hdr->runs = cpu_to_le64(endgrp->fdp.runs);
for (i = 0; i < nruh; i++) {
ruhd->ruht = NVME_RUHT_INITIALLY_ISOLATED;
ruhd++;
}
} else {
/* 1 bit for RUH in PIF -> 2 RUHs max. */
hdr->nrg = cpu_to_le16(1);
hdr->nruh = cpu_to_le16(1);
hdr->maxpids = cpu_to_le16(NVME_FDP_MAXPIDS - 1);
hdr->nnss = cpu_to_le32(1);
hdr->runs = cpu_to_le64(96 * MiB);
ruhd->ruht = NVME_RUHT_INITIALLY_ISOLATED;
}
return nvme_c2h(n, (uint8_t *)buf + off, trans_len, req);
}
static uint16_t nvme_fdp_ruh_usage(NvmeCtrl *n, uint32_t endgrpid,
uint32_t dw10, uint32_t dw12,
uint32_t buf_len, uint64_t off,
NvmeRequest *req)
{
NvmeRuHandle *ruh;
NvmeRuhuLog *hdr;
NvmeRuhuDescr *ruhud;
NvmeEnduranceGroup *endgrp;
g_autofree uint8_t *buf = NULL;
uint32_t log_size, trans_len;
uint16_t i;
if (endgrpid != 1 || !n->subsys) {
return NVME_INVALID_FIELD | NVME_DNR;
}
endgrp = &n->subsys->endgrp;
if (!endgrp->fdp.enabled) {
return NVME_FDP_DISABLED | NVME_DNR;
}
log_size = sizeof(NvmeRuhuLog) + endgrp->fdp.nruh * sizeof(NvmeRuhuDescr);
if (off >= log_size) {
return NVME_INVALID_FIELD | NVME_DNR;
}
trans_len = MIN(log_size - off, buf_len);
buf = g_malloc0(log_size);
hdr = (NvmeRuhuLog *)buf;
ruhud = (NvmeRuhuDescr *)(hdr + 1);
ruh = endgrp->fdp.ruhs;
hdr->nruh = cpu_to_le16(endgrp->fdp.nruh);
for (i = 0; i < endgrp->fdp.nruh; i++, ruhud++, ruh++) {
ruhud->ruha = ruh->ruha;
}
return nvme_c2h(n, (uint8_t *)buf + off, trans_len, req);
}
static uint16_t nvme_fdp_stats(NvmeCtrl *n, uint32_t endgrpid, uint32_t buf_len,
uint64_t off, NvmeRequest *req)
{
NvmeEnduranceGroup *endgrp;
NvmeFdpStatsLog log = {};
uint32_t trans_len;
if (off >= sizeof(NvmeFdpStatsLog)) {
return NVME_INVALID_FIELD | NVME_DNR;
}
if (endgrpid != 1 || !n->subsys) {
return NVME_INVALID_FIELD | NVME_DNR;
}
if (!n->subsys->endgrp.fdp.enabled) {
return NVME_FDP_DISABLED | NVME_DNR;
}
endgrp = &n->subsys->endgrp;
trans_len = MIN(sizeof(log) - off, buf_len);
/* spec value is 128 bit, we only use 64 bit */
log.hbmw[0] = cpu_to_le64(endgrp->fdp.hbmw);
log.mbmw[0] = cpu_to_le64(endgrp->fdp.mbmw);
log.mbe[0] = cpu_to_le64(endgrp->fdp.mbe);
return nvme_c2h(n, (uint8_t *)&log + off, trans_len, req);
}
static uint16_t nvme_fdp_events(NvmeCtrl *n, uint32_t endgrpid,
uint32_t buf_len, uint64_t off,
NvmeRequest *req)
{
NvmeEnduranceGroup *endgrp;
NvmeCmd *cmd = &req->cmd;
bool host_events = (cmd->cdw10 >> 8) & 0x1;
uint32_t log_size, trans_len;
NvmeFdpEventBuffer *ebuf;
g_autofree NvmeFdpEventsLog *elog = NULL;
NvmeFdpEvent *event;
if (endgrpid != 1 || !n->subsys) {
return NVME_INVALID_FIELD | NVME_DNR;
}
endgrp = &n->subsys->endgrp;
if (!endgrp->fdp.enabled) {
return NVME_FDP_DISABLED | NVME_DNR;
}
if (host_events) {
ebuf = &endgrp->fdp.host_events;
} else {
ebuf = &endgrp->fdp.ctrl_events;
}
log_size = sizeof(NvmeFdpEventsLog) + ebuf->nelems * sizeof(NvmeFdpEvent);
trans_len = MIN(log_size - off, buf_len);
elog = g_malloc0(log_size);
elog->num_events = cpu_to_le32(ebuf->nelems);
event = (NvmeFdpEvent *)(elog + 1);
if (ebuf->nelems && ebuf->start == ebuf->next) {
unsigned int nelems = (NVME_FDP_MAX_EVENTS - ebuf->start);
/* wrap over, copy [start;NVME_FDP_MAX_EVENTS[ and [0; next[ */
memcpy(event, &ebuf->events[ebuf->start],
sizeof(NvmeFdpEvent) * nelems);
memcpy(event + nelems, ebuf->events,
sizeof(NvmeFdpEvent) * ebuf->next);
} else if (ebuf->start < ebuf->next) {
memcpy(event, &ebuf->events[ebuf->start],
sizeof(NvmeFdpEvent) * (ebuf->next - ebuf->start));
}
return nvme_c2h(n, (uint8_t *)elog + off, trans_len, req);
}
static uint16_t nvme_get_log(NvmeCtrl *n, NvmeRequest *req)
{
NvmeCmd *cmd = &req->cmd;
@ -4635,13 +5120,14 @@ static uint16_t nvme_get_log(NvmeCtrl *n, NvmeRequest *req)
uint8_t lsp = (dw10 >> 8) & 0xf;
uint8_t rae = (dw10 >> 15) & 0x1;
uint8_t csi = le32_to_cpu(cmd->cdw14) >> 24;
uint32_t numdl, numdu;
uint32_t numdl, numdu, lspi;
uint64_t off, lpol, lpou;
size_t len;
uint16_t status;
numdl = (dw10 >> 16);
numdu = (dw11 & 0xffff);
lspi = (dw11 >> 16);
lpol = dw12;
lpou = dw13;
@ -4672,6 +5158,14 @@ static uint16_t nvme_get_log(NvmeCtrl *n, NvmeRequest *req)
return nvme_cmd_effects(n, csi, len, off, req);
case NVME_LOG_ENDGRP:
return nvme_endgrp_info(n, rae, len, off, req);
case NVME_LOG_FDP_CONFS:
return nvme_fdp_confs(n, lspi, len, off, req);
case NVME_LOG_FDP_RUH_USAGE:
return nvme_fdp_ruh_usage(n, lspi, dw10, dw12, len, off, req);
case NVME_LOG_FDP_STATS:
return nvme_fdp_stats(n, lspi, len, off, req);
case NVME_LOG_FDP_EVENTS:
return nvme_fdp_events(n, lspi, len, off, req);
default:
trace_pci_nvme_err_invalid_log_page(nvme_cid(req), lid);
return NVME_INVALID_FIELD | NVME_DNR;
@ -5258,6 +5752,84 @@ static uint16_t nvme_get_feature_timestamp(NvmeCtrl *n, NvmeRequest *req)
return nvme_c2h(n, (uint8_t *)&timestamp, sizeof(timestamp), req);
}
static int nvme_get_feature_fdp(NvmeCtrl *n, uint32_t endgrpid,
uint32_t *result)
{
*result = 0;
if (!n->subsys || !n->subsys->endgrp.fdp.enabled) {
return NVME_INVALID_FIELD | NVME_DNR;
}
*result = FIELD_DP16(0, FEAT_FDP, FDPE, 1);
*result = FIELD_DP16(*result, FEAT_FDP, CONF_NDX, 0);
return NVME_SUCCESS;
}
static uint16_t nvme_get_feature_fdp_events(NvmeCtrl *n, NvmeNamespace *ns,
NvmeRequest *req, uint32_t *result)
{
NvmeCmd *cmd = &req->cmd;
uint32_t cdw11 = le32_to_cpu(cmd->cdw11);
uint16_t ph = cdw11 & 0xffff;
uint8_t noet = (cdw11 >> 16) & 0xff;
uint16_t ruhid, ret;
uint32_t nentries = 0;
uint8_t s_events_ndx = 0;
size_t s_events_siz = sizeof(NvmeFdpEventDescr) * noet;
g_autofree NvmeFdpEventDescr *s_events = g_malloc0(s_events_siz);
NvmeRuHandle *ruh;
NvmeFdpEventDescr *s_event;
if (!n->subsys || !n->subsys->endgrp.fdp.enabled) {
return NVME_FDP_DISABLED | NVME_DNR;
}
if (!nvme_ph_valid(ns, ph)) {
return NVME_INVALID_FIELD | NVME_DNR;
}
ruhid = ns->fdp.phs[ph];
ruh = &n->subsys->endgrp.fdp.ruhs[ruhid];
assert(ruh);
if (unlikely(noet == 0)) {
return NVME_INVALID_FIELD | NVME_DNR;
}
for (uint8_t event_type = 0; event_type < FDP_EVT_MAX; event_type++) {
uint8_t shift = nvme_fdp_evf_shifts[event_type];
if (!shift && event_type) {
/*
* only first entry (event_type == 0) has a shift value of 0
* other entries are simply unpopulated.
*/
continue;
}
nentries++;
s_event = &s_events[s_events_ndx];
s_event->evt = event_type;
s_event->evta = (ruh->event_filter >> shift) & 0x1;
/* break if all `noet` entries are filled */
if ((++s_events_ndx) == noet) {
break;
}
}
ret = nvme_c2h(n, s_events, s_events_siz, req);
if (ret) {
return ret;
}
*result = nentries;
return NVME_SUCCESS;
}
static uint16_t nvme_get_feature(NvmeCtrl *n, NvmeRequest *req)
{
NvmeCmd *cmd = &req->cmd;
@ -5270,6 +5842,7 @@ static uint16_t nvme_get_feature(NvmeCtrl *n, NvmeRequest *req)
uint16_t iv;
NvmeNamespace *ns;
int i;
uint16_t endgrpid = 0, ret = NVME_SUCCESS;
static const uint32_t nvme_feature_default[NVME_FID_MAX] = {
[NVME_ARBITRATION] = NVME_ARB_AB_NOLIMIT,
@ -5367,6 +5940,33 @@ static uint16_t nvme_get_feature(NvmeCtrl *n, NvmeRequest *req)
case NVME_HOST_BEHAVIOR_SUPPORT:
return nvme_c2h(n, (uint8_t *)&n->features.hbs,
sizeof(n->features.hbs), req);
case NVME_FDP_MODE:
endgrpid = dw11 & 0xff;
if (endgrpid != 0x1) {
return NVME_INVALID_FIELD | NVME_DNR;
}
ret = nvme_get_feature_fdp(n, endgrpid, &result);
if (ret) {
return ret;
}
goto out;
case NVME_FDP_EVENTS:
if (!nvme_nsid_valid(n, nsid)) {
return NVME_INVALID_NSID | NVME_DNR;
}
ns = nvme_ns(n, nsid);
if (unlikely(!ns)) {
return NVME_INVALID_FIELD | NVME_DNR;
}
ret = nvme_get_feature_fdp_events(n, ns, req, &result);
if (ret) {
return ret;
}
goto out;
default:
break;
}
@ -5399,6 +5999,20 @@ defaults:
if (iv == n->admin_cq.vector) {
result |= NVME_INTVC_NOCOALESCING;
}
break;
case NVME_FDP_MODE:
endgrpid = dw11 & 0xff;
if (endgrpid != 0x1) {
return NVME_INVALID_FIELD | NVME_DNR;
}
ret = nvme_get_feature_fdp(n, endgrpid, &result);
if (ret) {
return ret;
}
goto out;
break;
default:
result = nvme_feature_default[fid];
@ -5407,7 +6021,7 @@ defaults:
out:
req->cqe.result = cpu_to_le32(result);
return NVME_SUCCESS;
return ret;
}
static uint16_t nvme_set_feature_timestamp(NvmeCtrl *n, NvmeRequest *req)
@ -5425,6 +6039,51 @@ static uint16_t nvme_set_feature_timestamp(NvmeCtrl *n, NvmeRequest *req)
return NVME_SUCCESS;
}
static uint16_t nvme_set_feature_fdp_events(NvmeCtrl *n, NvmeNamespace *ns,
NvmeRequest *req)
{
NvmeCmd *cmd = &req->cmd;
uint32_t cdw11 = le32_to_cpu(cmd->cdw11);
uint16_t ph = cdw11 & 0xffff;
uint8_t noet = (cdw11 >> 16) & 0xff;
uint16_t ret, ruhid;
uint8_t enable = le32_to_cpu(cmd->cdw12) & 0x1;
uint8_t event_mask = 0;
unsigned int i;
g_autofree uint8_t *events = g_malloc0(noet);
NvmeRuHandle *ruh = NULL;
assert(ns);
if (!n->subsys || !n->subsys->endgrp.fdp.enabled) {
return NVME_FDP_DISABLED | NVME_DNR;
}
if (!nvme_ph_valid(ns, ph)) {
return NVME_INVALID_FIELD | NVME_DNR;
}
ruhid = ns->fdp.phs[ph];
ruh = &n->subsys->endgrp.fdp.ruhs[ruhid];
ret = nvme_h2c(n, events, noet, req);
if (ret) {
return ret;
}
for (i = 0; i < noet; i++) {
event_mask |= (1 << nvme_fdp_evf_shifts[events[i]]);
}
if (enable) {
ruh->event_filter |= event_mask;
} else {
ruh->event_filter = ruh->event_filter & ~event_mask;
}
return NVME_SUCCESS;
}
static uint16_t nvme_set_feature(NvmeCtrl *n, NvmeRequest *req)
{
NvmeNamespace *ns = NULL;
@ -5584,6 +6243,11 @@ static uint16_t nvme_set_feature(NvmeCtrl *n, NvmeRequest *req)
return NVME_CMD_SET_CMB_REJECTED | NVME_DNR;
}
break;
case NVME_FDP_MODE:
/* spec: abort with cmd seq err if there's one or more NS' in endgrp */
return NVME_CMD_SEQ_ERROR | NVME_DNR;
case NVME_FDP_EVENTS:
return nvme_set_feature_fdp_events(n, ns, req);
default:
return NVME_FEAT_NOT_CHANGEABLE | NVME_DNR;
}
@ -6159,6 +6823,7 @@ static uint16_t nvme_directive_send(NvmeCtrl *n, NvmeRequest *req)
static uint16_t nvme_directive_receive(NvmeCtrl *n, NvmeRequest *req)
{
NvmeNamespace *ns;
uint32_t dw10 = le32_to_cpu(req->cmd.cdw10);
uint32_t dw11 = le32_to_cpu(req->cmd.cdw11);
uint32_t nsid = le32_to_cpu(req->cmd.nsid);
@ -6180,7 +6845,30 @@ static uint16_t nvme_directive_receive(NvmeCtrl *n, NvmeRequest *req)
return NVME_INVALID_FIELD | NVME_DNR;
}
return nvme_c2h(n, (uint8_t *)&id, trans_len, req);
ns = nvme_ns(n, nsid);
if (!ns) {
return NVME_INVALID_FIELD | NVME_DNR;
}
switch (dtype) {
case NVME_DIRECTIVE_IDENTIFY:
switch (doper) {
case NVME_DIRECTIVE_RETURN_PARAMS:
if (ns->endgrp->fdp.enabled) {
id.supported |= 1 << NVME_DIRECTIVE_DATA_PLACEMENT;
id.enabled |= 1 << NVME_DIRECTIVE_DATA_PLACEMENT;
id.persistent |= 1 << NVME_DIRECTIVE_DATA_PLACEMENT;
}
return nvme_c2h(n, (uint8_t *)&id, trans_len, req);
default:
return NVME_INVALID_FIELD | NVME_DNR;
}
default:
return NVME_INVALID_FIELD;
}
}
static uint16_t nvme_admin_cmd(NvmeCtrl *n, NvmeRequest *req)
@ -7545,6 +8233,10 @@ static void nvme_init_ctrl(NvmeCtrl *n, PCIDevice *pci_dev)
ctratt |= NVME_CTRATT_ENDGRPS;
id->endgidmax = cpu_to_le16(0x1);
if (n->subsys->endgrp.fdp.enabled) {
ctratt |= NVME_CTRATT_FDPS;
}
}
id->ctratt = cpu_to_le32(ctratt);