From: Chuck Lever <chuck.lever@xxxxxxxxxx>
Update the DMA error flow tracepoints to report the completion ID of
the failing context. This ties the wait/failure to a particular
operation or request, which is more useful than knowing only the
failing transport.
Signed-off-by: Chuck Lever <chuck.lever@xxxxxxxxxx>
---
include/trace/events/rpcrdma.h | 74 +++++++++++++++++++------------------
net/sunrpc/xprtrdma/svc_rdma_rw.c | 9 +++--
2 files changed, 42 insertions(+), 41 deletions(-)
diff --git a/include/trace/events/rpcrdma.h b/include/trace/events/rpcrdma.h
index f1c2022d39ca..bba758e5fb1d 100644
--- a/include/trace/events/rpcrdma.h
+++ b/include/trace/events/rpcrdma.h
@@ -1805,33 +1805,37 @@ DEFINE_SVC_DMA_EVENT(dma_unmap_page);
TRACE_EVENT(svcrdma_dma_map_rw_err,
TP_PROTO(
const struct svcxprt_rdma *rdma,
+ u64 offset,
+ u32 handle,
unsigned int nents,
int status
),
- TP_ARGS(rdma, nents, status),
+ TP_ARGS(rdma, offset, handle, nents, status),
TP_STRUCT__entry(
- __field(int, status)
+ __field(u32, cq_id)
+ __field(u32, handle)
+ __field(u64, offset)
__field(unsigned int, nents)
- __string(device, rdma->sc_cm_id->device->name)
- __string(addr, rdma->sc_xprt.xpt_remotebuf)
+ __field(int, status)
),
TP_fast_assign(
- __entry->status = status;
+ __entry->cq_id = rdma->sc_sq_cq->res.id;
+ __entry->handle = handle;
+ __entry->offset = offset;
__entry->nents = nents;
- __assign_str(device, rdma->sc_cm_id->device->name);
- __assign_str(addr, rdma->sc_xprt.xpt_remotebuf);
+ __entry->status = status;
),
- TP_printk("addr=%s device=%s nents=%u status=%d",
- __get_str(addr), __get_str(device), __entry->nents,
- __entry->status
+ TP_printk("cq.id=%u 0x%016llx:0x%08x nents=%u status=%d",
+ __entry->cq_id, (unsigned long long)__entry->offset,
+ __entry->handle, __entry->nents, __entry->status
)
);
-TRACE_EVENT(svcrdma_no_rwctx_err,
+TRACE_EVENT(svcrdma_rwctx_empty,
TP_PROTO(
const struct svcxprt_rdma *rdma,
unsigned int num_sges
@@ -1840,79 +1844,75 @@ TRACE_EVENT(svcrdma_no_rwctx_err,
TP_ARGS(rdma, num_sges),
TP_STRUCT__entry(
+ __field(u32, cq_id)
__field(unsigned int, num_sges)
- __string(device, rdma->sc_cm_id->device->name)
- __string(addr, rdma->sc_xprt.xpt_remotebuf)
),
TP_fast_assign(
+ __entry->cq_id = rdma->sc_sq_cq->res.id;
__entry->num_sges = num_sges;
- __assign_str(device, rdma->sc_cm_id->device->name);
- __assign_str(addr, rdma->sc_xprt.xpt_remotebuf);
),
- TP_printk("addr=%s device=%s num_sges=%d",
- __get_str(addr), __get_str(device), __entry->num_sges
+ TP_printk("cq.id=%u num_sges=%d",
+ __entry->cq_id, __entry->num_sges
)
);
TRACE_EVENT(svcrdma_page_overrun_err,
TP_PROTO(
- const struct svcxprt_rdma *rdma,
- const struct svc_rqst *rqst,
+ const struct rpc_rdma_cid *cid,
unsigned int pageno
),
- TP_ARGS(rdma, rqst, pageno),
+ TP_ARGS(cid, pageno),
TP_STRUCT__entry(
+ __field(u32, cq_id)
+ __field(int, completion_id)
__field(unsigned int, pageno)
- __field(u32, xid)
- __string(device, rdma->sc_cm_id->device->name)
- __string(addr, rdma->sc_xprt.xpt_remotebuf)
),
TP_fast_assign(
+ __entry->cq_id = cid->ci_queue_id;
+ __entry->completion_id = cid->ci_completion_id;
__entry->pageno = pageno;
- __entry->xid = __be32_to_cpu(rqst->rq_xid);
- __assign_str(device, rdma->sc_cm_id->device->name);
- __assign_str(addr, rdma->sc_xprt.xpt_remotebuf);
),
- TP_printk("addr=%s device=%s xid=0x%08x pageno=%u", __get_str(addr),
- __get_str(device), __entry->xid, __entry->pageno
+ TP_printk("cq.id=%u cid=%d pageno=%u",
+ __entry->cq_id, __entry->completion_id,
+ __entry->pageno
)
);
TRACE_EVENT(svcrdma_small_wrch_err,
TP_PROTO(
- const struct svcxprt_rdma *rdma,
+ const struct rpc_rdma_cid *cid,
unsigned int remaining,
unsigned int seg_no,
unsigned int num_segs
),
- TP_ARGS(rdma, remaining, seg_no, num_segs),
+ TP_ARGS(cid, remaining, seg_no, num_segs),
TP_STRUCT__entry(
+ __field(u32, cq_id)
+ __field(int, completion_id)
__field(unsigned int, remaining)
__field(unsigned int, seg_no)
__field(unsigned int, num_segs)
- __string(device, rdma->sc_cm_id->device->name)
- __string(addr, rdma->sc_xprt.xpt_remotebuf)
),
TP_fast_assign(
+ __entry->cq_id = cid->ci_queue_id;
+ __entry->completion_id = cid->ci_completion_id;
__entry->remaining = remaining;
__entry->seg_no = seg_no;
__entry->num_segs = num_segs;
- __assign_str(device, rdma->sc_cm_id->device->name);
- __assign_str(addr, rdma->sc_xprt.xpt_remotebuf);
),
- TP_printk("addr=%s device=%s remaining=%u seg_no=%u num_segs=%u",
- __get_str(addr), __get_str(device), __entry->remaining,
- __entry->seg_no, __entry->num_segs
+ TP_printk("cq.id=%u cid=%d remaining=%u seg_no=%u num_segs=%u",
+ __entry->cq_id, __entry->completion_id,
+ __entry->remaining, __entry->seg_no, __entry->num_segs
)
);
diff --git a/net/sunrpc/xprtrdma/svc_rdma_rw.c b/net/sunrpc/xprtrdma/svc_rdma_rw.c
index b06e49cc55fb..c06676714417 100644
--- a/net/sunrpc/xprtrdma/svc_rdma_rw.c
+++ b/net/sunrpc/xprtrdma/svc_rdma_rw.c
@@ -80,7 +80,7 @@ svc_rdma_get_rw_ctxt(struct svcxprt_rdma *rdma, unsigned int sges)
out_free:
kfree(ctxt);
out_noctx:
- trace_svcrdma_no_rwctx_err(rdma, sges);
+ trace_svcrdma_rwctx_empty(rdma, sges);
return NULL;
}
@@ -135,8 +135,9 @@ static int svc_rdma_rw_ctx_init(struct svcxprt_rdma *rdma,
ctxt->rw_sg_table.sgl, ctxt->rw_nents,
0, offset, handle, direction);
if (unlikely(ret < 0)) {
+ trace_svcrdma_dma_map_rw_err(rdma, offset, handle,
+ ctxt->rw_nents, ret);
svc_rdma_put_rw_ctxt(rdma, ctxt);
- trace_svcrdma_dma_map_rw_err(rdma, ctxt->rw_nents, ret);
}
return ret;
}
@@ -526,7 +527,7 @@ svc_rdma_build_writes(struct svc_rdma_write_info *info,
return 0;
out_overflow:
- trace_svcrdma_small_wrch_err(rdma, remaining, info->wi_seg_no,
+ trace_svcrdma_small_wrch_err(&cc->cc_cid, remaining, info->wi_seg_no,
info->wi_chunk->ch_segcount);
return -E2BIG;
}
@@ -766,7 +767,7 @@ static int svc_rdma_build_read_segment(struct svc_rdma_read_info *info,
return 0;
out_overrun:
- trace_svcrdma_page_overrun_err(cc->cc_rdma, rqstp, info->ri_pageno);
+ trace_svcrdma_page_overrun_err(&cc->cc_cid, info->ri_pageno);
return -EINVAL;
}
