| Message ID | 1465247706-4418-1-git-send-email-sudipm.mukherjee@gmail.com (mailing list archive) |
|---|---|
| State | Changes Requested |
| Headers | show |
On Mon, Jun 06, 2016 at 10:15:06PM +0100, Sudip Mukherjee wrote: > prot_sg_cnt has been assigned with the value of ret which we have > already checked to be non-zero so prot_sg_cnt can never be zero at this > point of the code and hence the else part can never execute. > And since we know prot_sg_cnt is non zero there is no use for the > if condition also. > > Signed-off-by: Sudip Mukherjee <sudip.mukherjee@codethink.co.uk> And what about other places? 428 if (prot_sg_cnt) 429 ctx->sig->sig_wr.prot = &ctx->sig->prot.sge; .... 455 if (prot_sg_cnt) 456 ib_mr_pool_put(qp, &qp->rdma_mrs, ctx->sig->prot.mr); .... > --- > drivers/infiniband/core/rw.c | 24 ++++++++++-------------- > 1 file changed, 10 insertions(+), 14 deletions(-) > > diff --git a/drivers/infiniband/core/rw.c b/drivers/infiniband/core/rw.c > index 1eb9b12..a829a14 100644 > --- a/drivers/infiniband/core/rw.c > +++ b/drivers/infiniband/core/rw.c > @@ -386,21 +386,17 @@ int rdma_rw_ctx_signature_init(struct rdma_rw_ctx *ctx, struct ib_qp *qp, > count += ret; > prev_wr = &ctx->sig->data.reg_wr.wr; > > - if (prot_sg_cnt) { > - ret = rdma_rw_init_one_mr(qp, port_num, &ctx->sig->prot, > - prot_sg, prot_sg_cnt, 0); > - if (ret < 0) > - goto out_destroy_data_mr; > - count += ret; > + ret = rdma_rw_init_one_mr(qp, port_num, &ctx->sig->prot, > + prot_sg, prot_sg_cnt, 0); > + if (ret < 0) > + goto out_destroy_data_mr; > + count += ret; > > - if (ctx->sig->prot.inv_wr.next) > - prev_wr->next = &ctx->sig->prot.inv_wr; > - else > - prev_wr->next = &ctx->sig->prot.reg_wr.wr; > - prev_wr = &ctx->sig->prot.reg_wr.wr; > - } else { > - ctx->sig->prot.mr = NULL; > - } > + if (ctx->sig->prot.inv_wr.next) > + prev_wr->next = &ctx->sig->prot.inv_wr; > + else > + prev_wr->next = &ctx->sig->prot.reg_wr.wr; > + prev_wr = &ctx->sig->prot.reg_wr.wr; > > ctx->sig->sig_mr = ib_mr_pool_get(qp, &qp->sig_mrs); > if (!ctx->sig->sig_mr) { > -- > 1.9.1 >
On 07/06/16 00:15, Sudip Mukherjee wrote: > prot_sg_cnt has been assigned with the value of ret which we have > already checked to be non-zero so prot_sg_cnt can never be zero at this > point of the code and hence the else part can never execute. > And since we know prot_sg_cnt is non zero there is no use for the > if condition also. > > Signed-off-by: Sudip Mukherjee <sudip.mukherjee@codethink.co.uk> > --- > drivers/infiniband/core/rw.c | 24 ++++++++++-------------- > 1 file changed, 10 insertions(+), 14 deletions(-) > > diff --git a/drivers/infiniband/core/rw.c b/drivers/infiniband/core/rw.c > index 1eb9b12..a829a14 100644 > --- a/drivers/infiniband/core/rw.c > +++ b/drivers/infiniband/core/rw.c > @@ -386,21 +386,17 @@ int rdma_rw_ctx_signature_init(struct rdma_rw_ctx *ctx, struct ib_qp *qp, > count += ret; > prev_wr = &ctx->sig->data.reg_wr.wr; > > - if (prot_sg_cnt) { > - ret = rdma_rw_init_one_mr(qp, port_num, &ctx->sig->prot, > - prot_sg, prot_sg_cnt, 0); > - if (ret < 0) > - goto out_destroy_data_mr; > - count += ret; > + ret = rdma_rw_init_one_mr(qp, port_num, &ctx->sig->prot, > + prot_sg, prot_sg_cnt, 0); > + if (ret < 0) > + goto out_destroy_data_mr; > + count += ret; > > - if (ctx->sig->prot.inv_wr.next) > - prev_wr->next = &ctx->sig->prot.inv_wr; > - else > - prev_wr->next = &ctx->sig->prot.reg_wr.wr; > - prev_wr = &ctx->sig->prot.reg_wr.wr; > - } else { > - ctx->sig->prot.mr = NULL; > - } > + if (ctx->sig->prot.inv_wr.next) > + prev_wr->next = &ctx->sig->prot.inv_wr; > + else > + prev_wr->next = &ctx->sig->prot.reg_wr.wr; > + prev_wr = &ctx->sig->prot.reg_wr.wr; > > ctx->sig->sig_mr = ib_mr_pool_get(qp, &qp->sig_mrs); > if (!ctx->sig->sig_mr) { > Actually it looks like the data-integrity insert/strip operation (where protection sg list does not exist) is broken. It looks that the protection scatterlist should be done only if prot_sg_count was provided... I don't have access to mlx5 devices at the moment (still waiting to get some...) -- To unsubscribe from this list: send the line "unsubscribe linux-rdma" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On Tue, Jun 07, 2016 at 03:41:04PM +0300, Sagi Grimberg wrote: > Actually it looks like the data-integrity insert/strip operation > (where protection sg list does not exist) is broken. > > It looks that the protection scatterlist should be done only if > prot_sg_count was provided... > > I don't have access to mlx5 devices at the moment (still waiting > to get some...) I've reported the brokenness in the existing iSER code when I started working on signature MR support in the generic API. As no one helped managed to figure out the issue or how it should operate I clone the existing semantics, even given I knew they are broken. -- To unsubscribe from this list: send the line "unsubscribe linux-rdma" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
>> Actually it looks like the data-integrity insert/strip operation >> (where protection sg list does not exist) is broken. >> >> It looks that the protection scatterlist should be done only if >> prot_sg_count was provided... >> >> I don't have access to mlx5 devices at the moment (still waiting >> to get some...) Hey Christoph, > I've reported the brokenness in the existing iSER code when I started > working on signature MR support in the generic API. As no one helped > managed to figure out the issue or how it should operate I clone the > existing semantics, even given I knew they are broken. There are two modes for the insert/strip operation, one is with PI on memory and without on the wire (which is useful when having a target working against legacy initiators), and one is with PI on the wire but not in memory. You reported that the latter was broken (the less useful case) and my understanding is that it was a result of the target<->isert interaction. The last time I used the first mode it worked fine. I'll try to get my hands on some mlx5 devices soon and have a look into this. I still want to change the API so it will hide all this noise from the consumer but I need to find some spare cycles for it. In the meantime, CC'ing Max from Mellanox and maybe he can check if PI insert/strip are still alive. Cheers, Sagi. -- To unsubscribe from this list: send the line "unsubscribe linux-rdma" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
diff --git a/drivers/infiniband/core/rw.c b/drivers/infiniband/core/rw.c index 1eb9b12..a829a14 100644 --- a/drivers/infiniband/core/rw.c +++ b/drivers/infiniband/core/rw.c @@ -386,21 +386,17 @@ int rdma_rw_ctx_signature_init(struct rdma_rw_ctx *ctx, struct ib_qp *qp, count += ret; prev_wr = &ctx->sig->data.reg_wr.wr; - if (prot_sg_cnt) { - ret = rdma_rw_init_one_mr(qp, port_num, &ctx->sig->prot, - prot_sg, prot_sg_cnt, 0); - if (ret < 0) - goto out_destroy_data_mr; - count += ret; + ret = rdma_rw_init_one_mr(qp, port_num, &ctx->sig->prot, + prot_sg, prot_sg_cnt, 0); + if (ret < 0) + goto out_destroy_data_mr; + count += ret; - if (ctx->sig->prot.inv_wr.next) - prev_wr->next = &ctx->sig->prot.inv_wr; - else - prev_wr->next = &ctx->sig->prot.reg_wr.wr; - prev_wr = &ctx->sig->prot.reg_wr.wr; - } else { - ctx->sig->prot.mr = NULL; - } + if (ctx->sig->prot.inv_wr.next) + prev_wr->next = &ctx->sig->prot.inv_wr; + else + prev_wr->next = &ctx->sig->prot.reg_wr.wr; + prev_wr = &ctx->sig->prot.reg_wr.wr; ctx->sig->sig_mr = ib_mr_pool_get(qp, &qp->sig_mrs); if (!ctx->sig->sig_mr) {
prot_sg_cnt has been assigned with the value of ret which we have already checked to be non-zero so prot_sg_cnt can never be zero at this point of the code and hence the else part can never execute. And since we know prot_sg_cnt is non zero there is no use for the if condition also. Signed-off-by: Sudip Mukherjee <sudip.mukherjee@codethink.co.uk> --- drivers/infiniband/core/rw.c | 24 ++++++++++-------------- 1 file changed, 10 insertions(+), 14 deletions(-)