int ring_fd;
struct io_u **io_u_index;
+ char *md_buf;
int *fds;
unsigned int uncached;
unsigned int nowait;
unsigned int force_async;
+ unsigned int md_per_io_size;
+ unsigned int pi_act;
+ unsigned int apptag;
+ unsigned int apptag_mask;
+ unsigned int prchk;
+ char *pi_chk;
enum uring_cmd_type cmd_type;
};
.group = FIO_OPT_G_IOURING,
},
CMDPRIO_OPTIONS(struct ioring_options, FIO_OPT_G_IOURING),
+ {
+ .name = "md_per_io_size",
+ .lname = "Separate Metadata Buffer Size per I/O",
+ .type = FIO_OPT_INT,
+ .off1 = offsetof(struct ioring_options, md_per_io_size),
+ .def = "0",
+ .help = "Size of separate metadata buffer per I/O (Default: 0)",
+ .category = FIO_OPT_C_ENGINE,
+ .group = FIO_OPT_G_IOURING,
+ },
+ {
+ .name = "pi_act",
+ .lname = "Protection Information Action",
+ .type = FIO_OPT_BOOL,
+ .off1 = offsetof(struct ioring_options, pi_act),
+ .def = "1",
+ .help = "Protection Information Action bit (pi_act=1 or pi_act=0)",
+ .category = FIO_OPT_C_ENGINE,
+ .group = FIO_OPT_G_IOURING,
+ },
+ {
+ .name = "pi_chk",
+ .lname = "Protection Information Check",
+ .type = FIO_OPT_STR_STORE,
+ .off1 = offsetof(struct ioring_options, pi_chk),
+ .def = NULL,
+ .help = "Control of Protection Information Checking (pi_chk=GUARD,REFTAG,APPTAG)",
+ .category = FIO_OPT_C_ENGINE,
+ .group = FIO_OPT_G_IOURING,
+ },
+ {
+ .name = "apptag",
+ .lname = "Application Tag used in Protection Information",
+ .type = FIO_OPT_INT,
+ .off1 = offsetof(struct ioring_options, apptag),
+ .def = "0x1234",
+ .help = "Application Tag used in Protection Information field (Default: 0x1234)",
+ .category = FIO_OPT_C_ENGINE,
+ .group = FIO_OPT_G_IOURING,
+ },
+ {
+ .name = "apptag_mask",
+ .lname = "Application Tag Mask",
+ .type = FIO_OPT_INT,
+ .off1 = offsetof(struct ioring_options, apptag_mask),
+ .def = "0xffff",
+ .help = "Application Tag Mask used with Application Tag (Default: 0xffff)",
+ .category = FIO_OPT_C_ENGINE,
+ .group = FIO_OPT_G_IOURING,
+ },
{
.name = NULL,
},
struct ioring_options *o = td->eo;
struct io_uring_cqe *cqe;
struct io_u *io_u;
+ struct nvme_data *data;
unsigned index;
+ int ret;
index = (event + ld->cq_ring_off) & ld->cq_ring_mask;
if (o->cmd_type == FIO_URING_CMD_NVME)
else
io_u->error = 0;
+ if (o->cmd_type == FIO_URING_CMD_NVME) {
+ data = FILE_ENG_DATA(io_u->file);
+ if (data->pi_type && (io_u->ddir == DDIR_READ) && !o->pi_act) {
+ ret = fio_nvme_pi_verify(data, io_u);
+ if (ret)
+ io_u->error = ret;
+ }
+ }
+
return io_u;
}
return r < 0 ? r : events;
}
+static inline void fio_ioring_cmd_nvme_pi(struct thread_data *td,
+ struct io_u *io_u)
+{
+ struct ioring_data *ld = td->io_ops_data;
+ struct ioring_options *o = td->eo;
+ struct nvme_uring_cmd *cmd;
+ struct io_uring_sqe *sqe;
+ struct nvme_cmd_ext_io_opts ext_opts = {0};
+ struct nvme_data *data = FILE_ENG_DATA(io_u->file);
+
+ if (io_u->ddir == DDIR_TRIM)
+ return;
+
+ sqe = &ld->sqes[(io_u->index) << 1];
+ cmd = (struct nvme_uring_cmd *)sqe->cmd;
+
+ if (data->pi_type) {
+ if (o->pi_act)
+ ext_opts.io_flags |= NVME_IO_PRINFO_PRACT;
+ ext_opts.io_flags |= o->prchk;
+ ext_opts.apptag = o->apptag;
+ ext_opts.apptag_mask = o->apptag_mask;
+ }
+
+ fio_nvme_pi_fill(cmd, io_u, &ext_opts);
+}
+
static inline void fio_ioring_cmdprio_prep(struct thread_data *td,
struct io_u *io_u)
{
struct io_u *io_u)
{
struct ioring_data *ld = td->io_ops_data;
+ struct ioring_options *o = td->eo;
struct io_sq_ring *ring = &ld->sq_ring;
unsigned tail, next_tail;
if (ld->cmdprio.mode != CMDPRIO_MODE_NONE)
fio_ioring_cmdprio_prep(td, io_u);
+ if (!strcmp(td->io_ops->name, "io_uring_cmd") &&
+ o->cmd_type == FIO_URING_CMD_NVME)
+ fio_ioring_cmd_nvme_pi(td, io_u);
+
ring->array[tail & ld->sq_ring_mask] = io_u->index;
atomic_store_release(ring->tail, next_tail);
fio_cmdprio_cleanup(&ld->cmdprio);
free(ld->io_u_index);
+ free(ld->md_buf);
free(ld->iovecs);
free(ld->fds);
free(ld->dsm);
return 0;
}
+static void parse_prchk_flags(struct ioring_options *o)
+{
+ if (!o->pi_chk)
+ return;
+
+ if (strstr(o->pi_chk, "GUARD") != NULL)
+ o->prchk = NVME_IO_PRINFO_PRCHK_GUARD;
+ if (strstr(o->pi_chk, "REFTAG") != NULL)
+ o->prchk |= NVME_IO_PRINFO_PRCHK_REF;
+ if (strstr(o->pi_chk, "APPTAG") != NULL)
+ o->prchk |= NVME_IO_PRINFO_PRCHK_APP;
+}
+
static int fio_ioring_init(struct thread_data *td)
{
struct ioring_options *o = td->eo;
struct ioring_data *ld;
+ unsigned long long md_size;
int ret;
/* sqthread submission requires registered files */
/* io_u index */
ld->io_u_index = calloc(td->o.iodepth, sizeof(struct io_u *));
+
+ /*
+ * metadata buffer for nvme command.
+ * We are only supporting iomem=malloc / mem=malloc as of now.
+ */
+ if (!strcmp(td->io_ops->name, "io_uring_cmd") &&
+ (o->cmd_type == FIO_URING_CMD_NVME) && o->md_per_io_size) {
+ md_size = (unsigned long long) o->md_per_io_size
+ * (unsigned long long) td->o.iodepth;
+ md_size += page_mask + td->o.mem_align;
+ if (td->o.mem_align && td->o.mem_align > page_size)
+ md_size += td->o.mem_align - page_size;
+ if (td->o.mem_type == MEM_MALLOC) {
+ ld->md_buf = malloc(md_size);
+ if (!ld->md_buf) {
+ free(ld);
+ return 1;
+ }
+ } else {
+ log_err("fio: Only iomem=malloc or mem=malloc is supported\n");
+ free(ld);
+ return 1;
+ }
+ }
+ parse_prchk_flags(o);
+
ld->iovecs = calloc(td->o.iodepth, sizeof(struct iovec));
td->io_ops_data = ld;
static int fio_ioring_io_u_init(struct thread_data *td, struct io_u *io_u)
{
struct ioring_data *ld = td->io_ops_data;
+ struct ioring_options *o = td->eo;
+ struct nvme_pi_data *pi_data;
+ char *p;
ld->io_u_index[io_u->index] = io_u;
+
+ if (!strcmp(td->io_ops->name, "io_uring_cmd")) {
+ p = PTR_ALIGN(ld->md_buf, page_mask) + td->o.mem_align;
+ p += o->md_per_io_size * io_u->index;
+ io_u->mmap_data = p;
+
+ if (!o->pi_act) {
+ pi_data = calloc(1, sizeof(*pi_data));
+ pi_data->io_flags |= o->prchk;
+ pi_data->apptag_mask = o->apptag_mask;
+ pi_data->apptag = o->apptag;
+ io_u->engine_data = pi_data;
+ }
+ }
+
return 0;
}
+static void fio_ioring_io_u_free(struct thread_data *td, struct io_u *io_u)
+{
+ struct ioring_options *o = td->eo;
+ struct nvme_pi *pi;
+
+ if (!strcmp(td->io_ops->name, "io_uring_cmd") &&
+ (o->cmd_type == FIO_URING_CMD_NVME)) {
+ pi = io_u->engine_data;
+ free(pi);
+ io_u->engine_data = NULL;
+ }
+}
+
static int fio_ioring_open_file(struct thread_data *td, struct fio_file *f)
{
struct ioring_data *ld = td->io_ops_data;
if (o->cmd_type == FIO_URING_CMD_NVME) {
struct nvme_data *data = NULL;
- unsigned int nsid, lba_size = 0;
- __u32 ms = 0;
+ unsigned int lba_size = 0;
__u64 nlba = 0;
int ret;
/* Store the namespace-id and lba size. */
data = FILE_ENG_DATA(f);
if (data == NULL) {
- ret = fio_nvme_get_info(f, &nsid, &lba_size, &ms, &nlba);
- if (ret)
- return ret;
-
data = calloc(1, sizeof(struct nvme_data));
- data->nsid = nsid;
- if (ms)
- data->lba_ext = lba_size + ms;
- else
- data->lba_shift = ilog2(lba_size);
+ ret = fio_nvme_get_info(f, &nlba, o->pi_act, data);
+ if (ret) {
+ free(data);
+ return ret;
+ }
FILE_SET_ENG_DATA(f, data);
}
- assert(data->lba_shift < 32);
- lba_size = data->lba_ext ? data->lba_ext : (1U << data->lba_shift);
+ lba_size = data->lba_ext ? data->lba_ext : data->lba_size;
for_each_rw_ddir(ddir) {
if (td->o.min_bs[ddir] % lba_size ||
td->o.max_bs[ddir] % lba_size) {
if (data->lba_ext)
- log_err("block size must be a multiple of "
- "(LBA data size + Metadata size)\n");
+ log_err("%s: block size must be a multiple of (LBA data size + Metadata size)\n",
+ f->file_name);
else
- log_err("block size must be a multiple of LBA data size\n");
+ log_err("%s: block size must be a multiple of LBA data size\n",
+ f->file_name);
+ td_verror(td, EINVAL, "fio_ioring_cmd_open_file");
+ return 1;
+ }
+ if (data->ms && !data->lba_ext && ddir != DDIR_TRIM &&
+ (o->md_per_io_size < ((td->o.max_bs[ddir] / data->lba_size) *
+ data->ms))) {
+ log_err("%s: md_per_io_size should be at least %llu bytes\n",
+ f->file_name,
+ ((td->o.max_bs[ddir] / data->lba_size) * data->ms));
+ td_verror(td, EINVAL, "fio_ioring_cmd_open_file");
return 1;
}
}
if (o->cmd_type == FIO_URING_CMD_NVME) {
struct nvme_data *data = NULL;
- unsigned int nsid, lba_size = 0;
- __u32 ms = 0;
__u64 nlba = 0;
int ret;
- ret = fio_nvme_get_info(f, &nsid, &lba_size, &ms, &nlba);
- if (ret)
- return ret;
-
data = calloc(1, sizeof(struct nvme_data));
- data->nsid = nsid;
- if (ms)
- data->lba_ext = lba_size + ms;
- else
- data->lba_shift = ilog2(lba_size);
+ ret = fio_nvme_get_info(f, &nlba, o->pi_act, data);
+ if (ret) {
+ free(data);
+ return ret;
+ }
- f->real_file_size = lba_size * nlba;
+ f->real_file_size = data->lba_size * nlba;
fio_file_set_size_known(f);
FILE_SET_ENG_DATA(f, data);
.init = fio_ioring_init,
.post_init = fio_ioring_cmd_post_init,
.io_u_init = fio_ioring_io_u_init,
+ .io_u_free = fio_ioring_io_u_free,
.prep = fio_ioring_cmd_prep,
.queue = fio_ioring_queue,
.commit = fio_ioring_commit,