From f079af7a117504b5b307b727858c972261047907 Mon Sep 17 00:00:00 2001 From: Vipul Pandya Date: Thu, 14 Mar 2013 05:08:58 +0000 Subject: RDMA/cxgb4: Add Support for Chelsio T5 adapter Adds support for Chelsio T5 adapter. Enables T5's Write Combining feature. Signed-off-by: Vipul Pandya Signed-off-by: David S. Miller --- drivers/infiniband/hw/cxgb4/qp.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'drivers/infiniband/hw/cxgb4/qp.c') diff --git a/drivers/infiniband/hw/cxgb4/qp.c b/drivers/infiniband/hw/cxgb4/qp.c index 17ba4f8bc12d..c46024409c4e 100644 --- a/drivers/infiniband/hw/cxgb4/qp.c +++ b/drivers/infiniband/hw/cxgb4/qp.c @@ -76,7 +76,7 @@ static void dealloc_sq(struct c4iw_rdev *rdev, struct t4_sq *sq) static int alloc_oc_sq(struct c4iw_rdev *rdev, struct t4_sq *sq) { - if (!ocqp_support || !t4_ocqp_supported()) + if (!ocqp_support || !ocqp_supported(&rdev->lldi)) return -ENOSYS; sq->dma_addr = c4iw_ocqp_pool_alloc(rdev, sq->memsize); if (!sq->dma_addr) -- cgit v1.2.3 From 3cbdb928e2ddd16649769c8597a3ebc06c7594fd Mon Sep 17 00:00:00 2001 From: Vipul Pandya Date: Thu, 14 Mar 2013 05:08:59 +0000 Subject: RDMA/cxgb4: Turn off db coalescing when RDMA QPs are in use. Signed-off-by: Vipul Pandya Signed-off-by: David S. Miller --- drivers/infiniband/hw/cxgb4/qp.c | 20 ++++++++++++++++---- 1 file changed, 16 insertions(+), 4 deletions(-) (limited to 'drivers/infiniband/hw/cxgb4/qp.c') diff --git a/drivers/infiniband/hw/cxgb4/qp.c b/drivers/infiniband/hw/cxgb4/qp.c index c46024409c4e..da4869f41e35 100644 --- a/drivers/infiniband/hw/cxgb4/qp.c +++ b/drivers/infiniband/hw/cxgb4/qp.c @@ -42,10 +42,17 @@ static int ocqp_support = 1; module_param(ocqp_support, int, 0644); MODULE_PARM_DESC(ocqp_support, "Support on-chip SQs (default=1)"); -int db_fc_threshold = 2000; +int db_fc_threshold = 1000; module_param(db_fc_threshold, int, 0644); -MODULE_PARM_DESC(db_fc_threshold, "QP count/threshold that triggers automatic " - "db flow control mode (default = 2000)"); +MODULE_PARM_DESC(db_fc_threshold, + "QP count/threshold that triggers" + " automatic db flow control mode (default = 1000)"); + +int db_coalescing_threshold; +module_param(db_coalescing_threshold, int, 0644); +MODULE_PARM_DESC(db_coalescing_threshold, + "QP count/threshold that triggers" + " disabling db coalescing (default = 0)"); static void set_state(struct c4iw_qp *qhp, enum c4iw_qp_state state) { @@ -1448,6 +1455,8 @@ int c4iw_destroy_qp(struct ib_qp *ib_qp) rhp->db_state = NORMAL; idr_for_each(&rhp->qpidr, enable_qp_db, NULL); } + if (rhp->qpcnt <= db_coalescing_threshold) + cxgb4_enable_db_coalescing(rhp->rdev.lldi.ports[0]); spin_unlock_irq(&rhp->lock); atomic_dec(&qhp->refcnt); wait_event(qhp->wait, !atomic_read(&qhp->refcnt)); @@ -1559,11 +1568,14 @@ struct ib_qp *c4iw_create_qp(struct ib_pd *pd, struct ib_qp_init_attr *attrs, spin_lock_irq(&rhp->lock); if (rhp->db_state != NORMAL) t4_disable_wq_db(&qhp->wq); - if (++rhp->qpcnt > db_fc_threshold && rhp->db_state == NORMAL) { + rhp->qpcnt++; + if (rhp->qpcnt > db_fc_threshold && rhp->db_state == NORMAL) { rhp->rdev.stats.db_state_transitions++; rhp->db_state = FLOW_CONTROL; idr_for_each(&rhp->qpidr, disable_qp_db, NULL); } + if (rhp->qpcnt > db_coalescing_threshold) + cxgb4_disable_db_coalescing(rhp->rdev.lldi.ports[0]); ret = insert_handle_nolock(rhp, &rhp->qpidr, qhp, qhp->wq.sq.qid); spin_unlock_irq(&rhp->lock); if (ret) -- cgit v1.2.3 From 80ccdd60512fc19fa87bf02876c59aeeb82fe4bc Mon Sep 17 00:00:00 2001 From: Vipul Pandya Date: Thu, 14 Mar 2013 05:09:00 +0000 Subject: RDMA/cxgb4: Add module_params to enable DB FC & Coalescing on T5 Both DB Flow-Control and DB Coalescing are disabled by default on T5 Signed-off-by: Vipul Pandya Signed-off-by: David S. Miller --- drivers/infiniband/hw/cxgb4/qp.c | 10 ++++++---- 1 file changed, 6 insertions(+), 4 deletions(-) (limited to 'drivers/infiniband/hw/cxgb4/qp.c') diff --git a/drivers/infiniband/hw/cxgb4/qp.c b/drivers/infiniband/hw/cxgb4/qp.c index da4869f41e35..28592d45809b 100644 --- a/drivers/infiniband/hw/cxgb4/qp.c +++ b/drivers/infiniband/hw/cxgb4/qp.c @@ -1455,8 +1455,9 @@ int c4iw_destroy_qp(struct ib_qp *ib_qp) rhp->db_state = NORMAL; idr_for_each(&rhp->qpidr, enable_qp_db, NULL); } - if (rhp->qpcnt <= db_coalescing_threshold) - cxgb4_enable_db_coalescing(rhp->rdev.lldi.ports[0]); + if (db_coalescing_threshold >= 0) + if (rhp->qpcnt <= db_coalescing_threshold) + cxgb4_enable_db_coalescing(rhp->rdev.lldi.ports[0]); spin_unlock_irq(&rhp->lock); atomic_dec(&qhp->refcnt); wait_event(qhp->wait, !atomic_read(&qhp->refcnt)); @@ -1574,8 +1575,9 @@ struct ib_qp *c4iw_create_qp(struct ib_pd *pd, struct ib_qp_init_attr *attrs, rhp->db_state = FLOW_CONTROL; idr_for_each(&rhp->qpidr, disable_qp_db, NULL); } - if (rhp->qpcnt > db_coalescing_threshold) - cxgb4_disable_db_coalescing(rhp->rdev.lldi.ports[0]); + if (db_coalescing_threshold >= 0) + if (rhp->qpcnt > db_coalescing_threshold) + cxgb4_disable_db_coalescing(rhp->rdev.lldi.ports[0]); ret = insert_handle_nolock(rhp, &rhp->qpidr, qhp, qhp->wq.sq.qid); spin_unlock_irq(&rhp->lock); if (ret) -- cgit v1.2.3 From 42b6a949903d28f59c95f4c71080aa8b41e3d1d1 Mon Sep 17 00:00:00 2001 From: Vipul Pandya Date: Thu, 14 Mar 2013 05:09:01 +0000 Subject: RDMA/cxgb4: Use DSGLs for fastreg and adapter memory writes for T5. It enables direct DMA by HW to memory region PBL arrays and fast register PBL arrays from host memory, vs the T4 way of passing these arrays in the WR itself. The result is lower latency for memory registration, and larger PBL array support for fast register operations. This patch also updates ULP_TX_MEM_WRITE command fields for T5. Ordering bit of ULP_TX_MEM_WRITE is at bit position 22 in T5 and at 23 in T4. Signed-off-by: Vipul Pandya Signed-off-by: David S. Miller --- drivers/infiniband/hw/cxgb4/qp.c | 76 ++++++++++++++++++++++++++++------------ 1 file changed, 53 insertions(+), 23 deletions(-) (limited to 'drivers/infiniband/hw/cxgb4/qp.c') diff --git a/drivers/infiniband/hw/cxgb4/qp.c b/drivers/infiniband/hw/cxgb4/qp.c index 28592d45809b..90833d701631 100644 --- a/drivers/infiniband/hw/cxgb4/qp.c +++ b/drivers/infiniband/hw/cxgb4/qp.c @@ -54,6 +54,10 @@ MODULE_PARM_DESC(db_coalescing_threshold, "QP count/threshold that triggers" " disabling db coalescing (default = 0)"); +static int max_fr_immd = T4_MAX_FR_IMMD; +module_param(max_fr_immd, int, 0644); +MODULE_PARM_DESC(max_fr_immd, "fastreg threshold for using DSGL instead of immedate"); + static void set_state(struct c4iw_qp *qhp, enum c4iw_qp_state state) { unsigned long flag; @@ -539,7 +543,7 @@ static int build_rdma_recv(struct c4iw_qp *qhp, union t4_recv_wr *wqe, } static int build_fastreg(struct t4_sq *sq, union t4_wr *wqe, - struct ib_send_wr *wr, u8 *len16) + struct ib_send_wr *wr, u8 *len16, u8 t5dev) { struct fw_ri_immd *imdp; @@ -561,28 +565,51 @@ static int build_fastreg(struct t4_sq *sq, union t4_wr *wqe, wqe->fr.va_hi = cpu_to_be32(wr->wr.fast_reg.iova_start >> 32); wqe->fr.va_lo_fbo = cpu_to_be32(wr->wr.fast_reg.iova_start & 0xffffffff); - WARN_ON(pbllen > T4_MAX_FR_IMMD); - imdp = (struct fw_ri_immd *)(&wqe->fr + 1); - imdp->op = FW_RI_DATA_IMMD; - imdp->r1 = 0; - imdp->r2 = 0; - imdp->immdlen = cpu_to_be32(pbllen); - p = (__be64 *)(imdp + 1); - rem = pbllen; - for (i = 0; i < wr->wr.fast_reg.page_list_len; i++) { - *p = cpu_to_be64((u64)wr->wr.fast_reg.page_list->page_list[i]); - rem -= sizeof *p; - if (++p == (__be64 *)&sq->queue[sq->size]) - p = (__be64 *)sq->queue; - } - BUG_ON(rem < 0); - while (rem) { - *p = 0; - rem -= sizeof *p; - if (++p == (__be64 *)&sq->queue[sq->size]) - p = (__be64 *)sq->queue; + + if (t5dev && use_dsgl && (pbllen > max_fr_immd)) { + struct c4iw_fr_page_list *c4pl = + to_c4iw_fr_page_list(wr->wr.fast_reg.page_list); + struct fw_ri_dsgl *sglp; + + for (i = 0; i < wr->wr.fast_reg.page_list_len; i++) { + wr->wr.fast_reg.page_list->page_list[i] = (__force u64) + cpu_to_be64((u64) + wr->wr.fast_reg.page_list->page_list[i]); + } + + sglp = (struct fw_ri_dsgl *)(&wqe->fr + 1); + sglp->op = FW_RI_DATA_DSGL; + sglp->r1 = 0; + sglp->nsge = cpu_to_be16(1); + sglp->addr0 = cpu_to_be64(c4pl->dma_addr); + sglp->len0 = cpu_to_be32(pbllen); + + *len16 = DIV_ROUND_UP(sizeof(wqe->fr) + sizeof(*sglp), 16); + } else { + imdp = (struct fw_ri_immd *)(&wqe->fr + 1); + imdp->op = FW_RI_DATA_IMMD; + imdp->r1 = 0; + imdp->r2 = 0; + imdp->immdlen = cpu_to_be32(pbllen); + p = (__be64 *)(imdp + 1); + rem = pbllen; + for (i = 0; i < wr->wr.fast_reg.page_list_len; i++) { + *p = cpu_to_be64( + (u64)wr->wr.fast_reg.page_list->page_list[i]); + rem -= sizeof(*p); + if (++p == (__be64 *)&sq->queue[sq->size]) + p = (__be64 *)sq->queue; + } + BUG_ON(rem < 0); + while (rem) { + *p = 0; + rem -= sizeof(*p); + if (++p == (__be64 *)&sq->queue[sq->size]) + p = (__be64 *)sq->queue; + } + *len16 = DIV_ROUND_UP(sizeof(wqe->fr) + sizeof(*imdp) + + pbllen, 16); } - *len16 = DIV_ROUND_UP(sizeof wqe->fr + sizeof *imdp + pbllen, 16); return 0; } @@ -683,7 +710,10 @@ int c4iw_post_send(struct ib_qp *ibqp, struct ib_send_wr *wr, case IB_WR_FAST_REG_MR: fw_opcode = FW_RI_FR_NSMR_WR; swsqe->opcode = FW_RI_FAST_REGISTER; - err = build_fastreg(&qhp->wq.sq, wqe, wr, &len16); + err = build_fastreg(&qhp->wq.sq, wqe, wr, &len16, + is_t5( + qhp->rhp->rdev.lldi.adapter_type) ? + 1 : 0); break; case IB_WR_LOCAL_INV: if (wr->send_flags & IB_SEND_FENCE) -- cgit v1.2.3 From 9919d5bd01b9eaf4928439e804dd70de24ea9637 Mon Sep 17 00:00:00 2001 From: Vipul Pandya Date: Thu, 14 Mar 2013 05:09:04 +0000 Subject: RDMA/cxgb4: Fix onchip queue support for T5 T5 adapter does not support onchip queue memory. Present logic fails to allocate QP for T5 and returns an error. Also, if module parameter ocqp_support is zero then we are unable to allocate QP which should not be the case. Ideally if ocqp_support parameter is 0 or onchip queue support is disable then host QP should be allocated before returning an error. Signed-off-by: Vipul Pandya Signed-off-by: David S. Miller --- drivers/infiniband/hw/cxgb4/qp.c | 13 +++++-------- 1 file changed, 5 insertions(+), 8 deletions(-) (limited to 'drivers/infiniband/hw/cxgb4/qp.c') diff --git a/drivers/infiniband/hw/cxgb4/qp.c b/drivers/infiniband/hw/cxgb4/qp.c index 90833d701631..9fe6f1e84373 100644 --- a/drivers/infiniband/hw/cxgb4/qp.c +++ b/drivers/infiniband/hw/cxgb4/qp.c @@ -140,7 +140,7 @@ static int create_qp(struct c4iw_rdev *rdev, struct t4_wq *wq, int wr_len; struct c4iw_wr_wait wr_wait; struct sk_buff *skb; - int ret; + int ret = 0; int eqsize; wq->sq.qid = c4iw_get_qpid(rdev, uctx); @@ -180,17 +180,14 @@ static int create_qp(struct c4iw_rdev *rdev, struct t4_wq *wq, } if (user) { - ret = alloc_oc_sq(rdev, &wq->sq); - if (ret) + if (alloc_oc_sq(rdev, &wq->sq) && alloc_host_sq(rdev, &wq->sq)) goto free_hwaddr; - - ret = alloc_host_sq(rdev, &wq->sq); - if (ret) - goto free_sq; - } else + } else { ret = alloc_host_sq(rdev, &wq->sq); if (ret) goto free_hwaddr; + } + memset(wq->sq.queue, 0, wq->sq.memsize); dma_unmap_addr_set(&wq->sq, mapping, wq->sq.dma_addr); -- cgit v1.2.3 From 55e57a780a10c9fd734603ec4b32644791ef5b05 Mon Sep 17 00:00:00 2001 From: Wei Yongjun Date: Fri, 15 Mar 2013 09:42:12 +0000 Subject: RDMA/cxgb4: Fix error return code in create_qp() Fix to return a negative error code from the error handling case instead of 0, as returned elsewhere in this function. Signed-off-by: Wei Yongjun Acked-by: Steve Wise Signed-off-by: Roland Dreier --- drivers/infiniband/hw/cxgb4/qp.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) (limited to 'drivers/infiniband/hw/cxgb4/qp.c') diff --git a/drivers/infiniband/hw/cxgb4/qp.c b/drivers/infiniband/hw/cxgb4/qp.c index 17ba4f8bc12d..70b1808a08f4 100644 --- a/drivers/infiniband/hw/cxgb4/qp.c +++ b/drivers/infiniband/hw/cxgb4/qp.c @@ -186,8 +186,10 @@ static int create_qp(struct c4iw_rdev *rdev, struct t4_wq *wq, wq->rq.queue = dma_alloc_coherent(&(rdev->lldi.pdev->dev), wq->rq.memsize, &(wq->rq.dma_addr), GFP_KERNEL); - if (!wq->rq.queue) + if (!wq->rq.queue) { + ret = -ENOMEM; goto free_sq; + } PDBG("%s sq base va 0x%p pa 0x%llx rq base va 0x%p pa 0x%llx\n", __func__, wq->sq.queue, (unsigned long long)virt_to_phys(wq->sq.queue), -- cgit v1.2.3 From 5b0c275926b8149c555da874bb4ec258ea3292aa Mon Sep 17 00:00:00 2001 From: Thadeu Lima de Souza Cascardo Date: Mon, 1 Apr 2013 20:13:39 +0000 Subject: RDMA/cxgb4: Fix SQ allocation when on-chip SQ is disabled Commit c079c28714e4 ("RDMA/cxgb4: Fix error handling in create_qp()") broke SQ allocation. Instead of falling back to host allocation when on-chip allocation fails, it tries to allocate both. And when it does, and we try to free the address from the genpool using the host address, we hit a BUG and the system crashes as below. We create a new function that has the previous behavior and properly propagate the error, as intended. kernel BUG at /usr/src/packages/BUILD/kernel-ppc64-3.0.68/linux-3.0/lib/genalloc.c:340! Oops: Exception in kernel mode, sig: 5 [#1] SMP NR_CPUS=1024 NUMA pSeries Modules linked in: rdma_ucm rdma_cm ib_addr ib_cm iw_cm ib_sa ib_mad ib_uverbs iw_cxgb4 ib_core ip6t_LOG xt_tcpudp xt_pkttype ipt_LOG xt_limit ip6t_REJECT nf_conntrack_ipv6 nf_defrag_ipv6 ip6table_raw xt_NOTRACK ipt_REJECT xt_state iptable_raw iptable_filter ip6table_mangle nf_conntrack_netbios_ns nf_conntrack_broadcast nf_conntrack_ipv4 nf_conntrack nf_defrag_ipv4 ip_tables ip6table_filter ip6_tables x_tables fuse loop dm_mod ipv6 ipv6_lib sr_mod cdrom ibmveth(X) cxgb4 sg ext3 jbd mbcache sd_mod crc_t10dif scsi_dh_emc scsi_dh_hp_sw scsi_dh_alua scsi_dh_rdac scsi_dh ibmvscsic(X) scsi_transport_srp scsi_tgt scsi_mod Supported: Yes NIP: c00000000037d41c LR: d000000003913824 CTR: c00000000037d3b0 REGS: c0000001f350ae50 TRAP: 0700 Tainted: G X (3.0.68-0.9-ppc64) MSR: 8000000000029032 CR: 24042482 XER: 00000001 TASK = c0000001f6f2a840[3616] 'rping' THREAD: c0000001f3508000 CPU: 0 GPR00: c0000001f6e875c8 c0000001f350b0d0 c000000000fc9690 c0000001f6e875c0 GPR04: 00000000000c0000 0000000000010000 0000000000000000 c0000000009d482a GPR08: 000000006a170000 0000000000100000 c0000001f350b140 c0000001f6e875c8 GPR12: d000000003915dd0 c000000003f40000 000000003e3ecfa8 c0000001f350bea0 GPR16: c0000001f350bcd0 00000000003c0000 0000000000040100 c0000001f6e74a80 GPR20: d00000000399a898 c0000001f6e74ac8 c0000001fad91600 c0000001f6e74ab0 GPR24: c0000001f7d23f80 0000000000000000 0000000000000002 000000006a170000 GPR28: 000000000000000c c0000001f584c8d0 d000000003925180 c0000001f6e875c8 NIP [c00000000037d41c] .gen_pool_free+0x6c/0xf8 LR [d000000003913824] .c4iw_ocqp_pool_free+0x8c/0xd8 [iw_cxgb4] Call Trace: [c0000001f350b0d0] [c0000001f350b180] 0xc0000001f350b180 (unreliable) [c0000001f350b170] [d000000003913824] .c4iw_ocqp_pool_free+0x8c/0xd8 [iw_cxgb4] [c0000001f350b210] [d00000000390fd70] .dealloc_sq+0x90/0xb0 [iw_cxgb4] [c0000001f350b280] [d00000000390fe08] .destroy_qp+0x78/0xf8 [iw_cxgb4] [c0000001f350b310] [d000000003912738] .c4iw_destroy_qp+0x208/0x2d0 [iw_cxgb4] [c0000001f350b460] [d000000003861874] .ib_destroy_qp+0x5c/0x130 [ib_core] [c0000001f350b510] [d0000000039911bc] .ib_uverbs_cleanup_ucontext+0x174/0x4f8 [ib_uverbs] [c0000001f350b5f0] [d000000003991568] .ib_uverbs_close+0x28/0x70 [ib_uverbs] [c0000001f350b670] [c0000000001e7b2c] .__fput+0xdc/0x278 [c0000001f350b720] [c0000000001a9590] .remove_vma+0x68/0xd8 [c0000001f350b7b0] [c0000000001a9720] .exit_mmap+0x120/0x160 [c0000001f350b8d0] [c0000000000af330] .mmput+0x80/0x160 [c0000001f350b960] [c0000000000b5d0c] .exit_mm+0x1ac/0x1e8 [c0000001f350ba10] [c0000000000b8154] .do_exit+0x1b4/0x4b8 [c0000001f350bad0] [c0000000000b84b0] .do_group_exit+0x58/0xf8 [c0000001f350bb60] [c0000000000ce9f4] .get_signal_to_deliver+0x2f4/0x5d0 [c0000001f350bc60] [c000000000017ee4] .do_signal_pending+0x6c/0x3e0 [c0000001f350bdb0] [c0000000000182cc] .do_signal+0x74/0x78 [c0000001f350be30] [c000000000009e74] do_work+0x24/0x28 Signed-off-by: Thadeu Lima de Souza Cascardo Cc: Emil Goode Cc: Acked-by: Steve Wise Signed-off-by: Roland Dreier --- drivers/infiniband/hw/cxgb4/qp.c | 25 +++++++++++++------------ 1 file changed, 13 insertions(+), 12 deletions(-) (limited to 'drivers/infiniband/hw/cxgb4/qp.c') diff --git a/drivers/infiniband/hw/cxgb4/qp.c b/drivers/infiniband/hw/cxgb4/qp.c index 70b1808a08f4..ed49ab345b6e 100644 --- a/drivers/infiniband/hw/cxgb4/qp.c +++ b/drivers/infiniband/hw/cxgb4/qp.c @@ -100,6 +100,16 @@ static int alloc_host_sq(struct c4iw_rdev *rdev, struct t4_sq *sq) return 0; } +static int alloc_sq(struct c4iw_rdev *rdev, struct t4_sq *sq, int user) +{ + int ret = -ENOSYS; + if (user) + ret = alloc_oc_sq(rdev, sq); + if (ret) + ret = alloc_host_sq(rdev, sq); + return ret; +} + static int destroy_qp(struct c4iw_rdev *rdev, struct t4_wq *wq, struct c4iw_dev_ucontext *uctx) { @@ -168,18 +178,9 @@ static int create_qp(struct c4iw_rdev *rdev, struct t4_wq *wq, goto free_sw_rq; } - if (user) { - ret = alloc_oc_sq(rdev, &wq->sq); - if (ret) - goto free_hwaddr; - - ret = alloc_host_sq(rdev, &wq->sq); - if (ret) - goto free_sq; - } else - ret = alloc_host_sq(rdev, &wq->sq); - if (ret) - goto free_hwaddr; + ret = alloc_sq(rdev, &wq->sq, user); + if (ret) + goto free_hwaddr; memset(wq->sq.queue, 0, wq->sq.memsize); dma_unmap_addr_set(&wq->sq, mapping, wq->sq.dma_addr); -- cgit v1.2.3