Commit 2563e2f3 authored by Max Gurtovoy's avatar Max Gurtovoy Committed by Jason Gunthorpe

RDMA/mlx5: Use PA mapping for PI handover

If possibe, avoid doing a UMR operation to register data and protection
buffers (via MTT/KLM mkeys). Instead, use the local DMA key and map the
SG lists using PA access. This is safe, since the internal key for data
and protection never exposed to the remote server (only signature key
might be exposed). If PA mappings are not possible, perform mapping
using MTT/KLM descriptors.

The setup of the tested benchmark (using iSER ULP):
 - 2 servers with 24 cores (1 initiator and 1 target)
 - ConnectX-4/ConnectX-5 adapters
 - 24 target sessions with 1 LUN each
 - ramdisk backstore
 - PI active

Performance results running fio (24 jobs, 128 iodepth) using
write_generate=1 and read_verify=1 (w/w.o patch):

bs      IOPS(read)        IOPS(write)
----    ----------        ----------
512   1266.4K/1262.4K    1720.1K/1732.1K
4k    793139/570902      1129.6K/773982
32k   72660/72086        97229/96164

Using write_generate=0 and read_verify=0 (w/w.o patch):
bs      IOPS(read)        IOPS(write)
----    ----------        ----------
512   1590.2K/1600.1K    1828.2K/1830.3K
4k    1078.1K/937272     1142.1K/815304
32k   77012/77369        98125/97435
Signed-off-by: default avatarMax Gurtovoy <maxg@mellanox.com>
Signed-off-by: default avatarIsrael Rukshin <israelr@mellanox.com>
Suggested-by: default avatarSagi Grimberg <sagi@grimberg.me>
Signed-off-by: default avatarJason Gunthorpe <jgg@mellanox.com>
parent de0ae958
...@@ -609,6 +609,7 @@ struct mlx5_ib_mr { ...@@ -609,6 +609,7 @@ struct mlx5_ib_mr {
struct mlx5_ib_mr *pi_mr; struct mlx5_ib_mr *pi_mr;
struct mlx5_ib_mr *klm_mr; struct mlx5_ib_mr *klm_mr;
struct mlx5_ib_mr *mtt_mr; struct mlx5_ib_mr *mtt_mr;
u64 data_iova;
u64 pi_iova; u64 pi_iova;
atomic_t num_leaf_free; atomic_t num_leaf_free;
......
...@@ -2001,6 +2001,40 @@ int mlx5_ib_check_mr_status(struct ib_mr *ibmr, u32 check_mask, ...@@ -2001,6 +2001,40 @@ int mlx5_ib_check_mr_status(struct ib_mr *ibmr, u32 check_mask,
return ret; return ret;
} }
static int
mlx5_ib_map_pa_mr_sg_pi(struct ib_mr *ibmr, struct scatterlist *data_sg,
int data_sg_nents, unsigned int *data_sg_offset,
struct scatterlist *meta_sg, int meta_sg_nents,
unsigned int *meta_sg_offset)
{
struct mlx5_ib_mr *mr = to_mmr(ibmr);
unsigned int sg_offset = 0;
int n = 0;
mr->meta_length = 0;
if (data_sg_nents == 1) {
n++;
mr->ndescs = 1;
if (data_sg_offset)
sg_offset = *data_sg_offset;
mr->data_length = sg_dma_len(data_sg) - sg_offset;
mr->data_iova = sg_dma_address(data_sg) + sg_offset;
if (meta_sg_nents == 1) {
n++;
mr->meta_ndescs = 1;
if (meta_sg_offset)
sg_offset = *meta_sg_offset;
else
sg_offset = 0;
mr->meta_length = sg_dma_len(meta_sg) - sg_offset;
mr->pi_iova = sg_dma_address(meta_sg) + sg_offset;
}
ibmr->length = mr->data_length + mr->meta_length;
}
return n;
}
static int static int
mlx5_ib_sg_to_klms(struct mlx5_ib_mr *mr, mlx5_ib_sg_to_klms(struct mlx5_ib_mr *mr,
struct scatterlist *sgl, struct scatterlist *sgl,
...@@ -2099,7 +2133,6 @@ mlx5_ib_map_mtt_mr_sg_pi(struct ib_mr *ibmr, struct scatterlist *data_sg, ...@@ -2099,7 +2133,6 @@ mlx5_ib_map_mtt_mr_sg_pi(struct ib_mr *ibmr, struct scatterlist *data_sg,
struct mlx5_ib_mr *mr = to_mmr(ibmr); struct mlx5_ib_mr *mr = to_mmr(ibmr);
struct mlx5_ib_mr *pi_mr = mr->mtt_mr; struct mlx5_ib_mr *pi_mr = mr->mtt_mr;
int n; int n;
u64 iova;
pi_mr->ndescs = 0; pi_mr->ndescs = 0;
pi_mr->meta_ndescs = 0; pi_mr->meta_ndescs = 0;
...@@ -2115,13 +2148,14 @@ mlx5_ib_map_mtt_mr_sg_pi(struct ib_mr *ibmr, struct scatterlist *data_sg, ...@@ -2115,13 +2148,14 @@ mlx5_ib_map_mtt_mr_sg_pi(struct ib_mr *ibmr, struct scatterlist *data_sg,
if (n != data_sg_nents) if (n != data_sg_nents)
return n; return n;
iova = pi_mr->ibmr.iova; pi_mr->data_iova = pi_mr->ibmr.iova;
pi_mr->data_length = pi_mr->ibmr.length; pi_mr->data_length = pi_mr->ibmr.length;
pi_mr->ibmr.length = pi_mr->data_length; pi_mr->ibmr.length = pi_mr->data_length;
ibmr->length = pi_mr->data_length; ibmr->length = pi_mr->data_length;
if (meta_sg_nents) { if (meta_sg_nents) {
u64 page_mask = ~((u64)ibmr->page_size - 1); u64 page_mask = ~((u64)ibmr->page_size - 1);
u64 iova = pi_mr->data_iova;
n += ib_sg_to_pages(&pi_mr->ibmr, meta_sg, meta_sg_nents, n += ib_sg_to_pages(&pi_mr->ibmr, meta_sg, meta_sg_nents,
meta_sg_offset, mlx5_set_page_pi); meta_sg_offset, mlx5_set_page_pi);
...@@ -2181,6 +2215,7 @@ mlx5_ib_map_klm_mr_sg_pi(struct ib_mr *ibmr, struct scatterlist *data_sg, ...@@ -2181,6 +2215,7 @@ mlx5_ib_map_klm_mr_sg_pi(struct ib_mr *ibmr, struct scatterlist *data_sg,
DMA_TO_DEVICE); DMA_TO_DEVICE);
/* This is zero-based memory region */ /* This is zero-based memory region */
pi_mr->data_iova = 0;
pi_mr->ibmr.iova = 0; pi_mr->ibmr.iova = 0;
pi_mr->pi_iova = pi_mr->data_length; pi_mr->pi_iova = pi_mr->data_length;
ibmr->length = pi_mr->ibmr.length; ibmr->length = pi_mr->ibmr.length;
...@@ -2194,11 +2229,27 @@ int mlx5_ib_map_mr_sg_pi(struct ib_mr *ibmr, struct scatterlist *data_sg, ...@@ -2194,11 +2229,27 @@ int mlx5_ib_map_mr_sg_pi(struct ib_mr *ibmr, struct scatterlist *data_sg,
unsigned int *meta_sg_offset) unsigned int *meta_sg_offset)
{ {
struct mlx5_ib_mr *mr = to_mmr(ibmr); struct mlx5_ib_mr *mr = to_mmr(ibmr);
struct mlx5_ib_mr *pi_mr = mr->mtt_mr; struct mlx5_ib_mr *pi_mr = NULL;
int n; int n;
WARN_ON(ibmr->type != IB_MR_TYPE_INTEGRITY); WARN_ON(ibmr->type != IB_MR_TYPE_INTEGRITY);
mr->ndescs = 0;
mr->data_length = 0;
mr->data_iova = 0;
mr->meta_ndescs = 0;
mr->pi_iova = 0;
/*
* As a performance optimization, if possible, there is no need to
* perform UMR operation to register the data/metadata buffers.
* First try to map the sg lists to PA descriptors with local_dma_lkey.
* Fallback to UMR only in case of a failure.
*/
n = mlx5_ib_map_pa_mr_sg_pi(ibmr, data_sg, data_sg_nents,
data_sg_offset, meta_sg, meta_sg_nents,
meta_sg_offset);
if (n == data_sg_nents + meta_sg_nents)
goto out;
/* /*
* As a performance optimization, if possible, there is no need to map * As a performance optimization, if possible, there is no need to map
* the sg lists to KLM descriptors. First try to map the sg lists to MTT * the sg lists to KLM descriptors. First try to map the sg lists to MTT
...@@ -2207,6 +2258,7 @@ int mlx5_ib_map_mr_sg_pi(struct ib_mr *ibmr, struct scatterlist *data_sg, ...@@ -2207,6 +2258,7 @@ int mlx5_ib_map_mr_sg_pi(struct ib_mr *ibmr, struct scatterlist *data_sg,
* (especially in high load). * (especially in high load).
* Use KLM (indirect access) only if it's mandatory. * Use KLM (indirect access) only if it's mandatory.
*/ */
pi_mr = mr->mtt_mr;
n = mlx5_ib_map_mtt_mr_sg_pi(ibmr, data_sg, data_sg_nents, n = mlx5_ib_map_mtt_mr_sg_pi(ibmr, data_sg, data_sg_nents,
data_sg_offset, meta_sg, meta_sg_nents, data_sg_offset, meta_sg, meta_sg_nents,
meta_sg_offset); meta_sg_offset);
...@@ -2224,7 +2276,10 @@ int mlx5_ib_map_mr_sg_pi(struct ib_mr *ibmr, struct scatterlist *data_sg, ...@@ -2224,7 +2276,10 @@ int mlx5_ib_map_mr_sg_pi(struct ib_mr *ibmr, struct scatterlist *data_sg,
/* This is zero-based memory region */ /* This is zero-based memory region */
ibmr->iova = 0; ibmr->iova = 0;
mr->pi_mr = pi_mr; mr->pi_mr = pi_mr;
if (pi_mr)
ibmr->sig_attrs->meta_length = pi_mr->meta_length; ibmr->sig_attrs->meta_length = pi_mr->meta_length;
else
ibmr->sig_attrs->meta_length = mr->meta_length;
return 0; return 0;
} }
......
...@@ -4562,7 +4562,7 @@ static int set_sig_data_segment(const struct ib_send_wr *send_wr, ...@@ -4562,7 +4562,7 @@ static int set_sig_data_segment(const struct ib_send_wr *send_wr,
data_len = pi_mr->data_length; data_len = pi_mr->data_length;
data_key = pi_mr->ibmr.lkey; data_key = pi_mr->ibmr.lkey;
data_va = pi_mr->ibmr.iova; data_va = pi_mr->data_iova;
if (pi_mr->meta_ndescs) { if (pi_mr->meta_ndescs) {
prot_len = pi_mr->meta_length; prot_len = pi_mr->meta_length;
prot_key = pi_mr->ibmr.lkey; prot_key = pi_mr->ibmr.lkey;
...@@ -4912,6 +4912,7 @@ static int _mlx5_ib_post_send(struct ib_qp *ibqp, const struct ib_send_wr *wr, ...@@ -4912,6 +4912,7 @@ static int _mlx5_ib_post_send(struct ib_qp *ibqp, const struct ib_send_wr *wr,
struct mlx5_ib_qp *qp; struct mlx5_ib_qp *qp;
struct mlx5_ib_mr *mr; struct mlx5_ib_mr *mr;
struct mlx5_ib_mr *pi_mr; struct mlx5_ib_mr *pi_mr;
struct mlx5_ib_mr pa_pi_mr;
struct ib_sig_attrs *sig_attrs; struct ib_sig_attrs *sig_attrs;
struct mlx5_wqe_xrc_seg *xrc; struct mlx5_wqe_xrc_seg *xrc;
struct mlx5_bf *bf; struct mlx5_bf *bf;
...@@ -5026,36 +5027,63 @@ static int _mlx5_ib_post_send(struct ib_qp *ibqp, const struct ib_send_wr *wr, ...@@ -5026,36 +5027,63 @@ static int _mlx5_ib_post_send(struct ib_qp *ibqp, const struct ib_send_wr *wr,
break; break;
case IB_WR_REG_MR_INTEGRITY: case IB_WR_REG_MR_INTEGRITY:
memset(&reg_pi_wr, 0, sizeof(struct ib_reg_wr)); qp->sq.wr_data[idx] = IB_WR_REG_MR_INTEGRITY;
mr = to_mmr(reg_wr(wr)->mr); mr = to_mmr(reg_wr(wr)->mr);
pi_mr = mr->pi_mr; pi_mr = mr->pi_mr;
if (pi_mr) {
memset(&reg_pi_wr, 0,
sizeof(struct ib_reg_wr));
reg_pi_wr.mr = &pi_mr->ibmr; reg_pi_wr.mr = &pi_mr->ibmr;
reg_pi_wr.access = reg_wr(wr)->access; reg_pi_wr.access = reg_wr(wr)->access;
reg_pi_wr.key = pi_mr->ibmr.rkey; reg_pi_wr.key = pi_mr->ibmr.rkey;
qp->sq.wr_data[idx] = IB_WR_REG_MR_INTEGRITY;
ctrl->imm = cpu_to_be32(reg_pi_wr.key); ctrl->imm = cpu_to_be32(reg_pi_wr.key);
/* UMR for data + protection registration */ /* UMR for data + prot registration */
err = set_reg_wr(qp, &reg_pi_wr, &seg, &size, err = set_reg_wr(qp, &reg_pi_wr, &seg,
&cur_edge, false); &size, &cur_edge,
false);
if (err) { if (err) {
*bad_wr = wr; *bad_wr = wr;
goto out; goto out;
} }
finish_wqe(qp, ctrl, seg, size, cur_edge, idx, finish_wqe(qp, ctrl, seg, size,
wr->wr_id, nreq, fence, cur_edge, idx, wr->wr_id,
nreq, fence,
MLX5_OPCODE_UMR); MLX5_OPCODE_UMR);
err = begin_wqe(qp, &seg, &ctrl, wr, &idx, err = begin_wqe(qp, &seg, &ctrl, wr,
&size, &cur_edge, nreq); &idx, &size, &cur_edge,
nreq);
if (err) { if (err) {
mlx5_ib_warn(dev, "\n"); mlx5_ib_warn(dev, "\n");
err = -ENOMEM; err = -ENOMEM;
*bad_wr = wr; *bad_wr = wr;
goto out; goto out;
} }
} else {
memset(&pa_pi_mr, 0,
sizeof(struct mlx5_ib_mr));
/* No UMR, use local_dma_lkey */
pa_pi_mr.ibmr.lkey =
mr->ibmr.pd->local_dma_lkey;
pa_pi_mr.ndescs = mr->ndescs;
pa_pi_mr.data_length = mr->data_length;
pa_pi_mr.data_iova = mr->data_iova;
if (mr->meta_ndescs) {
pa_pi_mr.meta_ndescs =
mr->meta_ndescs;
pa_pi_mr.meta_length =
mr->meta_length;
pa_pi_mr.pi_iova = mr->pi_iova;
}
pa_pi_mr.ibmr.length = mr->ibmr.length;
mr->pi_mr = &pa_pi_mr;
}
ctrl->imm = cpu_to_be32(mr->ibmr.rkey); ctrl->imm = cpu_to_be32(mr->ibmr.rkey);
/* UMR for sig MR */ /* UMR for sig MR */
err = set_pi_umr_wr(wr, qp, &seg, &size, err = set_pi_umr_wr(wr, qp, &seg, &size,
......
Markdown is supported
0%
or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment