* [Qemu-devel] [PATCH v4 0/9] Misc fixes to pvrdma device
@ 2019-03-03 20:33 Yuval Shaia
2019-03-03 20:33 ` [Qemu-devel] [PATCH v4 1/9] hw/rdma: Switch to generic error reporting way Yuval Shaia
` (8 more replies)
0 siblings, 9 replies; 17+ messages in thread
From: Yuval Shaia @ 2019-03-03 20:33 UTC (permalink / raw)
To: dgilbert, yuval.shaia, marcel.apfelbaum, armbru, qemu-devel
Hi,
Please review the following patch-set which consist of cosmetics fixes to
device's user interface (traces, error_report and monitor) and some bug
fixes.
Thanks Markus, Eric, Marcel and David for reviewing v0.
David, please see version notes below, since i restructured the HMP part
your second review is needed, thanks in advance.
v0 -> v1:
* Explain why device attributes are exposed only in HMP interface.
* Squash the 3 patches related to HMP interface into one.
* Make monitor dump function simple.
* Make HMP interface available only if pvrdma is included (detected by
build robot).
* Remove patch 03/10 ("Warn when too many consecutive poll CQ triggered
on an empty CQ) and add the two counters to patch 0/7 (monitor).
* Add Marcel's R-Bs.
* Add mutex protection to cqe_ctx list.
* Add two new patches.
v1 -> v2:
* Rename locked-lists to protected-lists in patch 2 and patch 6.
* Add Marcel's R-Bs.
v2 -> v3:
* Address some 32 bit host compilation issues.
v3 -> v4:
* Per suggestion from Markus rebase HMP report on
object_child_foreach_recursive().
* Strip off David's r-b from HMP patch (#4) because of the above.
Yuval Shaia (9):
hw/rdma: Switch to generic error reporting way
hw/rdma: Introduce protected qlist
hw/rdma: Protect against concurrent execution of poll_cq
{hmp,hw/pvrdma}: Expose device internals via monitor interface
hw/rdma: Free all MAD receive buffers when device is closed
hw/rdma: Free all receive buffers when QP is destroyed
hw/pvrdma: Delete unneeded function argument
hw/pvrdma: Delete pvrdma_exit function
hw/pvrdma: Unregister from shutdown notifier when device goes down
hmp-commands-info.hx | 14 +
hmp.c | 27 ++
hmp.h | 1 +
hw/rdma/Makefile.objs | 2 +-
hw/rdma/rdma_backend.c | 483 +++++++++++++++++-----------------
hw/rdma/rdma_backend.h | 3 +-
hw/rdma/rdma_backend_defs.h | 10 +-
hw/rdma/rdma_hmp.c | 30 +++
hw/rdma/rdma_rm.c | 193 ++++++++------
hw/rdma/rdma_rm.h | 7 +-
hw/rdma/rdma_rm_defs.h | 28 +-
hw/rdma/rdma_utils.c | 83 +++++-
hw/rdma/rdma_utils.h | 61 ++---
hw/rdma/trace-events | 32 ++-
hw/rdma/vmw/pvrdma.h | 7 +-
hw/rdma/vmw/pvrdma_cmd.c | 113 ++------
hw/rdma/vmw/pvrdma_dev_ring.c | 26 +-
hw/rdma/vmw/pvrdma_main.c | 161 +++++-------
hw/rdma/vmw/pvrdma_qp_ops.c | 52 +---
hw/rdma/vmw/trace-events | 16 +-
include/hw/rdma/rdma_hmp.h | 40 +++
21 files changed, 779 insertions(+), 610 deletions(-)
create mode 100644 hw/rdma/rdma_hmp.c
create mode 100644 include/hw/rdma/rdma_hmp.h
--
2.17.2
^ permalink raw reply [flat|nested] 17+ messages in thread
* [Qemu-devel] [PATCH v4 1/9] hw/rdma: Switch to generic error reporting way
2019-03-03 20:33 [Qemu-devel] [PATCH v4 0/9] Misc fixes to pvrdma device Yuval Shaia
@ 2019-03-03 20:33 ` Yuval Shaia
2019-03-03 20:33 ` [Qemu-devel] [PATCH v4 2/9] hw/rdma: Introduce protected qlist Yuval Shaia
` (7 subsequent siblings)
8 siblings, 0 replies; 17+ messages in thread
From: Yuval Shaia @ 2019-03-03 20:33 UTC (permalink / raw)
To: dgilbert, yuval.shaia, marcel.apfelbaum, armbru, qemu-devel
Utilize error_report for all pr_err calls and some pr_dbg that are
considered as errors.
For the remaining pr_dbg calls, the important ones were replaced by
trace points while other deleted.
Signed-off-by: Yuval Shaia <yuval.shaia@oracle.com>
Reviewed-by: Marcel Apfelbaum <marcel.apfelbaum@gmail.com>
---
hw/rdma/rdma_backend.c | 336 ++++++++++++++--------------------
hw/rdma/rdma_rm.c | 127 ++++++-------
hw/rdma/rdma_rm.h | 6 +-
hw/rdma/rdma_utils.c | 15 +-
hw/rdma/rdma_utils.h | 45 +----
hw/rdma/trace-events | 32 +++-
hw/rdma/vmw/pvrdma.h | 2 +-
hw/rdma/vmw/pvrdma_cmd.c | 113 +++---------
hw/rdma/vmw/pvrdma_dev_ring.c | 26 +--
hw/rdma/vmw/pvrdma_main.c | 132 +++++--------
hw/rdma/vmw/pvrdma_qp_ops.c | 49 ++---
hw/rdma/vmw/trace-events | 16 +-
12 files changed, 343 insertions(+), 556 deletions(-)
diff --git a/hw/rdma/rdma_backend.c b/hw/rdma/rdma_backend.c
index fd571f21e5..24bac00a20 100644
--- a/hw/rdma/rdma_backend.c
+++ b/hw/rdma/rdma_backend.c
@@ -14,7 +14,6 @@
*/
#include "qemu/osdep.h"
-#include "qemu/error-report.h"
#include "sysemu/sysemu.h"
#include "qapi/error.h"
#include "qapi/qmp/qlist.h"
@@ -39,7 +38,6 @@
typedef struct BackendCtx {
void *up_ctx;
- bool is_tx_req;
struct ibv_sge sge; /* Used to save MAD recv buffer */
} BackendCtx;
@@ -52,7 +50,7 @@ static void (*comp_handler)(void *ctx, struct ibv_wc *wc);
static void dummy_comp_handler(void *ctx, struct ibv_wc *wc)
{
- pr_err("No completion handler is registered\n");
+ rdma_error_report("No completion handler is registered");
}
static inline void complete_work(enum ibv_wc_status status, uint32_t vendor_err,
@@ -66,29 +64,24 @@ static inline void complete_work(enum ibv_wc_status status, uint32_t vendor_err,
comp_handler(ctx, &wc);
}
-static void poll_cq(RdmaDeviceResources *rdma_dev_res, struct ibv_cq *ibcq)
+static void rdma_poll_cq(RdmaDeviceResources *rdma_dev_res, struct ibv_cq *ibcq)
{
int i, ne;
BackendCtx *bctx;
struct ibv_wc wc[2];
- pr_dbg("Entering poll_cq loop on cq %p\n", ibcq);
do {
ne = ibv_poll_cq(ibcq, ARRAY_SIZE(wc), wc);
- pr_dbg("Got %d completion(s) from cq %p\n", ne, ibcq);
+ trace_rdma_poll_cq(ne, ibcq);
for (i = 0; i < ne; i++) {
- pr_dbg("wr_id=0x%" PRIx64 "\n", wc[i].wr_id);
- pr_dbg("status=%d\n", wc[i].status);
-
bctx = rdma_rm_get_cqe_ctx(rdma_dev_res, wc[i].wr_id);
if (unlikely(!bctx)) {
- pr_dbg("Error: Failed to find ctx for req %" PRId64 "\n",
- wc[i].wr_id);
+ rdma_error_report("No matching ctx for req %"PRId64,
+ wc[i].wr_id);
continue;
}
- pr_dbg("Processing %s CQE\n", bctx->is_tx_req ? "send" : "recv");
comp_handler(bctx->up_ctx, &wc[i]);
@@ -98,7 +91,7 @@ static void poll_cq(RdmaDeviceResources *rdma_dev_res, struct ibv_cq *ibcq)
} while (ne > 0);
if (ne < 0) {
- pr_dbg("Got error %d from ibv_poll_cq\n", ne);
+ rdma_error_report("ibv_poll_cq fail, rc=%d, errno=%d", ne, errno);
}
}
@@ -115,12 +108,10 @@ static void *comp_handler_thread(void *arg)
flags = fcntl(backend_dev->channel->fd, F_GETFL);
rc = fcntl(backend_dev->channel->fd, F_SETFL, flags | O_NONBLOCK);
if (rc < 0) {
- pr_dbg("Fail to change to non-blocking mode\n");
+ rdma_error_report("Failed to change backend channel FD to non-blocking");
return NULL;
}
- pr_dbg("Starting\n");
-
pfds[0].fd = backend_dev->channel->fd;
pfds[0].events = G_IO_IN | G_IO_HUP | G_IO_ERR;
@@ -132,27 +123,25 @@ static void *comp_handler_thread(void *arg)
} while (!rc && backend_dev->comp_thread.run);
if (backend_dev->comp_thread.run) {
- pr_dbg("Waiting for completion on channel %p\n", backend_dev->channel);
rc = ibv_get_cq_event(backend_dev->channel, &ev_cq, &ev_ctx);
- pr_dbg("ibv_get_cq_event=%d\n", rc);
if (unlikely(rc)) {
- pr_dbg("---> ibv_get_cq_event (%d)\n", rc);
+ rdma_error_report("ibv_get_cq_event fail, rc=%d, errno=%d", rc,
+ errno);
continue;
}
rc = ibv_req_notify_cq(ev_cq, 0);
if (unlikely(rc)) {
- pr_dbg("Error %d from ibv_req_notify_cq\n", rc);
+ rdma_error_report("ibv_req_notify_cq fail, rc=%d, errno=%d", rc,
+ errno);
}
- poll_cq(backend_dev->rdma_dev_res, ev_cq);
+ rdma_poll_cq(backend_dev->rdma_dev_res, ev_cq);
ibv_ack_cq_events(ev_cq, 1);
}
}
- pr_dbg("Going down\n");
-
/* TODO: Post cqe for all remaining buffs that were posted */
backend_dev->comp_thread.is_running = false;
@@ -177,55 +166,54 @@ static inline int rdmacm_mux_can_process_async(RdmaBackendDev *backend_dev)
return atomic_read(&backend_dev->rdmacm_mux.can_receive);
}
-static int check_mux_op_status(CharBackend *mad_chr_be)
+static int rdmacm_mux_check_op_status(CharBackend *mad_chr_be)
{
RdmaCmMuxMsg msg = {};
int ret;
- pr_dbg("Reading response\n");
ret = qemu_chr_fe_read_all(mad_chr_be, (uint8_t *)&msg, sizeof(msg));
if (ret != sizeof(msg)) {
- pr_dbg("Invalid message size %d, expecting %ld\n", ret, sizeof(msg));
+ rdma_error_report("Got invalid message from mux: size %d, expecting %d",
+ ret, (int)sizeof(msg));
return -EIO;
}
- pr_dbg("msg_type=%d\n", msg.hdr.msg_type);
- pr_dbg("op_code=%d\n", msg.hdr.op_code);
- pr_dbg("err_code=%d\n", msg.hdr.err_code);
+ trace_rdmacm_mux_check_op_status(msg.hdr.msg_type, msg.hdr.op_code,
+ msg.hdr.err_code);
if (msg.hdr.msg_type != RDMACM_MUX_MSG_TYPE_RESP) {
- pr_dbg("Invalid message type %d\n", msg.hdr.msg_type);
+ rdma_error_report("Got invalid message type %d", msg.hdr.msg_type);
return -EIO;
}
if (msg.hdr.err_code != RDMACM_MUX_ERR_CODE_OK) {
- pr_dbg("Operation failed in mux, error code %d\n", msg.hdr.err_code);
+ rdma_error_report("Operation failed in mux, error code %d",
+ msg.hdr.err_code);
return -EIO;
}
return 0;
}
-static int exec_rdmacm_mux_req(RdmaBackendDev *backend_dev, RdmaCmMuxMsg *msg)
+static int rdmacm_mux_send(RdmaBackendDev *backend_dev, RdmaCmMuxMsg *msg)
{
int rc = 0;
- pr_dbg("Executing request %d\n", msg->hdr.op_code);
-
msg->hdr.msg_type = RDMACM_MUX_MSG_TYPE_REQ;
+ trace_rdmacm_mux("send", msg->hdr.msg_type, msg->hdr.op_code);
disable_rdmacm_mux_async(backend_dev);
rc = qemu_chr_fe_write(backend_dev->rdmacm_mux.chr_be,
(const uint8_t *)msg, sizeof(*msg));
if (rc != sizeof(*msg)) {
enable_rdmacm_mux_async(backend_dev);
- pr_dbg("Fail to send request to rdmacm_mux (rc=%d)\n", rc);
+ rdma_error_report("Failed to send request to rdmacm_mux (rc=%d)", rc);
return -EIO;
}
- rc = check_mux_op_status(backend_dev->rdmacm_mux.chr_be);
+ rc = rdmacm_mux_check_op_status(backend_dev->rdmacm_mux.chr_be);
if (rc) {
- pr_dbg("Fail to execute rdmacm_mux request %d (rc=%d)\n",
- msg->hdr.op_code, rc);
+ rdma_error_report("Failed to execute rdmacm_mux request %d (rc=%d)",
+ msg->hdr.op_code, rc);
}
enable_rdmacm_mux_async(backend_dev);
@@ -237,7 +225,6 @@ static void stop_backend_thread(RdmaBackendThread *thread)
{
thread->run = false;
while (thread->is_running) {
- pr_dbg("Waiting for thread to complete\n");
sleep(THR_POLL_TO / SCALE_US / 2);
}
}
@@ -273,7 +260,7 @@ int rdma_backend_query_port(RdmaBackendDev *backend_dev,
rc = ibv_query_port(backend_dev->context, backend_dev->port_num, port_attr);
if (rc) {
- pr_dbg("Error %d from ibv_query_port\n", rc);
+ rdma_error_report("ibv_query_port fail, rc=%d, errno=%d", rc, errno);
return -EIO;
}
@@ -282,7 +269,7 @@ int rdma_backend_query_port(RdmaBackendDev *backend_dev,
void rdma_backend_poll_cq(RdmaDeviceResources *rdma_dev_res, RdmaBackendCQ *cq)
{
- poll_cq(rdma_dev_res, cq->ibcq);
+ rdma_poll_cq(rdma_dev_res, cq->ibcq);
}
static GHashTable *ah_hash;
@@ -294,8 +281,8 @@ static struct ibv_ah *create_ah(RdmaBackendDev *backend_dev, struct ibv_pd *pd,
struct ibv_ah *ah = g_hash_table_lookup(ah_hash, ah_key);
if (ah) {
- trace_create_ah_cache_hit(be64_to_cpu(dgid->global.subnet_prefix),
- be64_to_cpu(dgid->global.interface_id));
+ trace_rdma_create_ah_cache_hit(be64_to_cpu(dgid->global.subnet_prefix),
+ be64_to_cpu(dgid->global.interface_id));
g_bytes_unref(ah_key);
} else {
struct ibv_ah_attr ah_attr = {
@@ -312,13 +299,13 @@ static struct ibv_ah *create_ah(RdmaBackendDev *backend_dev, struct ibv_pd *pd,
g_hash_table_insert(ah_hash, ah_key, ah);
} else {
g_bytes_unref(ah_key);
- pr_dbg("Fail to create AH for gid <0x%" PRIx64 ", 0x%" PRIx64 ">\n",
- be64_to_cpu(dgid->global.subnet_prefix),
- be64_to_cpu(dgid->global.interface_id));
+ rdma_error_report("Failed to create AH for gid <0x%" PRIx64", 0x%"PRIx64">",
+ be64_to_cpu(dgid->global.subnet_prefix),
+ be64_to_cpu(dgid->global.interface_id));
}
- trace_create_ah_cache_miss(be64_to_cpu(dgid->global.subnet_prefix),
- be64_to_cpu(dgid->global.interface_id));
+ trace_rdma_create_ah_cache_miss(be64_to_cpu(dgid->global.subnet_prefix),
+ be64_to_cpu(dgid->global.interface_id));
}
return ah;
@@ -349,12 +336,10 @@ static int build_host_sge_array(RdmaDeviceResources *rdma_dev_res,
RdmaRmMR *mr;
int ssge_idx;
- pr_dbg("num_sge=%d\n", num_sge);
-
for (ssge_idx = 0; ssge_idx < num_sge; ssge_idx++) {
mr = rdma_rm_get_mr(rdma_dev_res, ssge[ssge_idx].lkey);
if (unlikely(!mr)) {
- pr_dbg("Invalid lkey 0x%x\n", ssge[ssge_idx].lkey);
+ rdma_error_report("Invalid lkey 0x%x", ssge[ssge_idx].lkey);
return VENDOR_ERR_INVLKEY | ssge[ssge_idx].lkey;
}
@@ -362,17 +347,28 @@ static int build_host_sge_array(RdmaDeviceResources *rdma_dev_res,
dsge->length = ssge[ssge_idx].length;
dsge->lkey = rdma_backend_mr_lkey(&mr->backend_mr);
- pr_dbg("ssge->addr=0x%" PRIx64 "\n", ssge[ssge_idx].addr);
- pr_dbg("dsge->addr=0x%" PRIx64 "\n", dsge->addr);
- pr_dbg("dsge->length=%d\n", dsge->length);
- pr_dbg("dsge->lkey=0x%x\n", dsge->lkey);
-
dsge++;
}
return 0;
}
+static void trace_mad_message(const char *title, char *buf, int len)
+{
+ int i;
+ char *b = g_malloc0(len * 3 + 1);
+ char b1[4];
+
+ for (i = 0; i < len; i++) {
+ sprintf(b1, "%.2X ", buf[i] & 0x000000FF);
+ strcat(b, b1);
+ }
+
+ trace_rdma_mad_message(title, len, b);
+
+ g_free(b);
+}
+
static int mad_send(RdmaBackendDev *backend_dev, uint8_t sgid_idx,
union ibv_gid *sgid, struct ibv_sge *sge, uint32_t num_sge)
{
@@ -380,8 +376,6 @@ static int mad_send(RdmaBackendDev *backend_dev, uint8_t sgid_idx,
char *hdr, *data;
int ret;
- pr_dbg("num_sge=%d\n", num_sge);
-
if (num_sge != 2) {
return -EINVAL;
}
@@ -390,7 +384,6 @@ static int mad_send(RdmaBackendDev *backend_dev, uint8_t sgid_idx,
memcpy(msg.hdr.sgid.raw, sgid->raw, sizeof(msg.hdr.sgid));
msg.umad_len = sge[0].length + sge[1].length;
- pr_dbg("umad_len=%d\n", msg.umad_len);
if (msg.umad_len > sizeof(msg.umad.mad)) {
return -ENOMEM;
@@ -398,36 +391,31 @@ static int mad_send(RdmaBackendDev *backend_dev, uint8_t sgid_idx,
msg.umad.hdr.addr.qpn = htobe32(1);
msg.umad.hdr.addr.grh_present = 1;
- pr_dbg("sgid_idx=%d\n", sgid_idx);
- pr_dbg("sgid=0x%llx\n", sgid->global.interface_id);
msg.umad.hdr.addr.gid_index = sgid_idx;
memcpy(msg.umad.hdr.addr.gid, sgid->raw, sizeof(msg.umad.hdr.addr.gid));
msg.umad.hdr.addr.hop_limit = 0xFF;
hdr = rdma_pci_dma_map(backend_dev->dev, sge[0].addr, sge[0].length);
if (!hdr) {
- pr_dbg("Fail to map to sge[0]\n");
return -ENOMEM;
}
data = rdma_pci_dma_map(backend_dev->dev, sge[1].addr, sge[1].length);
if (!data) {
- pr_dbg("Fail to map to sge[1]\n");
rdma_pci_dma_unmap(backend_dev->dev, hdr, sge[0].length);
return -ENOMEM;
}
- pr_dbg_buf("mad_hdr", hdr, sge[0].length);
- pr_dbg_buf("mad_data", data, sge[1].length);
-
memcpy(&msg.umad.mad[0], hdr, sge[0].length);
memcpy(&msg.umad.mad[sge[0].length], data, sge[1].length);
rdma_pci_dma_unmap(backend_dev->dev, data, sge[1].length);
rdma_pci_dma_unmap(backend_dev->dev, hdr, sge[0].length);
- ret = exec_rdmacm_mux_req(backend_dev, &msg);
+ trace_mad_message("send", msg.umad.mad, msg.umad_len);
+
+ ret = rdmacm_mux_send(backend_dev, &msg);
if (ret) {
- pr_dbg("Fail to send MAD to rdma_umadmux (%d)\n", ret);
+ rdma_error_report("Failed to send MAD to rdma_umadmux (%d)", ret);
return -EIO;
}
@@ -447,12 +435,11 @@ void rdma_backend_post_send(RdmaBackendDev *backend_dev,
int rc;
struct ibv_send_wr wr = {0}, *bad_wr;
- if (!qp->ibqp) { /* This field does not get initialized for QP0 and QP1 */
+ if (!qp->ibqp) { /* This field is not initialized for QP0 and QP1 */
if (qp_type == IBV_QPT_SMI) {
- pr_dbg("QP0 unsupported\n");
+ rdma_error_report("Got QP0 request");
complete_work(IBV_WC_GENERAL_ERR, VENDOR_ERR_QP0, ctx);
} else if (qp_type == IBV_QPT_GSI) {
- pr_dbg("QP1\n");
rc = mad_send(backend_dev, sgid_idx, sgid, sge, num_sge);
if (rc) {
complete_work(IBV_WC_GENERAL_ERR, VENDOR_ERR_MAD_SEND, ctx);
@@ -463,22 +450,17 @@ void rdma_backend_post_send(RdmaBackendDev *backend_dev,
return;
}
- pr_dbg("num_sge=%d\n", num_sge);
-
bctx = g_malloc0(sizeof(*bctx));
bctx->up_ctx = ctx;
- bctx->is_tx_req = 1;
rc = rdma_rm_alloc_cqe_ctx(backend_dev->rdma_dev_res, &bctx_id, bctx);
if (unlikely(rc)) {
- pr_dbg("Failed to allocate cqe_ctx\n");
complete_work(IBV_WC_GENERAL_ERR, VENDOR_ERR_NOMEM, ctx);
goto out_free_bctx;
}
rc = build_host_sge_array(backend_dev->rdma_dev_res, new_sge, sge, num_sge);
if (rc) {
- pr_dbg("Error: Failed to build host SGE array\n");
complete_work(IBV_WC_GENERAL_ERR, rc, ctx);
goto out_dealloc_cqe_ctx;
}
@@ -500,10 +482,9 @@ void rdma_backend_post_send(RdmaBackendDev *backend_dev,
wr.wr_id = bctx_id;
rc = ibv_post_send(qp->ibqp, &wr, &bad_wr);
- pr_dbg("ibv_post_send=%d\n", rc);
if (rc) {
- pr_dbg("Fail (%d, %d) to post send WQE to qpn %d\n", rc, errno,
- qp->ibqp->qp_num);
+ rdma_error_report("ibv_post_send fail, qpn=0x%x, rc=%d, errno=%d",
+ qp->ibqp->qp_num, rc, errno);
complete_work(IBV_WC_GENERAL_ERR, VENDOR_ERR_FAIL_BACKEND, ctx);
goto out_dealloc_cqe_ctx;
}
@@ -526,29 +507,23 @@ static unsigned int save_mad_recv_buffer(RdmaBackendDev *backend_dev,
uint32_t bctx_id;
if (num_sge != 1) {
- pr_dbg("Invalid num_sge (%d), expecting 1\n", num_sge);
+ rdma_error_report("Invalid num_sge (%d), expecting 1", num_sge);
return VENDOR_ERR_INV_NUM_SGE;
}
if (sge[0].length < RDMA_MAX_PRIVATE_DATA + sizeof(struct ibv_grh)) {
- pr_dbg("Too small buffer for MAD\n");
+ rdma_error_report("Too small buffer for MAD");
return VENDOR_ERR_INV_MAD_BUFF;
}
- pr_dbg("addr=0x%" PRIx64"\n", sge[0].addr);
- pr_dbg("length=%d\n", sge[0].length);
- pr_dbg("lkey=%d\n", sge[0].lkey);
-
bctx = g_malloc0(sizeof(*bctx));
rc = rdma_rm_alloc_cqe_ctx(backend_dev->rdma_dev_res, &bctx_id, bctx);
if (unlikely(rc)) {
g_free(bctx);
- pr_dbg("Fail to allocate cqe_ctx\n");
return VENDOR_ERR_NOMEM;
}
- pr_dbg("bctx_id %d, bctx %p, ctx %p\n", bctx_id, bctx, ctx);
bctx->up_ctx = ctx;
bctx->sge = *sge;
@@ -572,11 +547,10 @@ void rdma_backend_post_recv(RdmaBackendDev *backend_dev,
if (!qp->ibqp) { /* This field does not get initialized for QP0 and QP1 */
if (qp_type == IBV_QPT_SMI) {
- pr_dbg("QP0 unsupported\n");
+ rdma_error_report("Got QP0 request");
complete_work(IBV_WC_GENERAL_ERR, VENDOR_ERR_QP0, ctx);
}
if (qp_type == IBV_QPT_GSI) {
- pr_dbg("QP1\n");
rc = save_mad_recv_buffer(backend_dev, sge, num_sge, ctx);
if (rc) {
complete_work(IBV_WC_GENERAL_ERR, rc, ctx);
@@ -585,22 +559,17 @@ void rdma_backend_post_recv(RdmaBackendDev *backend_dev,
return;
}
- pr_dbg("num_sge=%d\n", num_sge);
-
bctx = g_malloc0(sizeof(*bctx));
bctx->up_ctx = ctx;
- bctx->is_tx_req = 0;
rc = rdma_rm_alloc_cqe_ctx(rdma_dev_res, &bctx_id, bctx);
if (unlikely(rc)) {
- pr_dbg("Failed to allocate cqe_ctx\n");
complete_work(IBV_WC_GENERAL_ERR, VENDOR_ERR_NOMEM, ctx);
goto out_free_bctx;
}
rc = build_host_sge_array(rdma_dev_res, new_sge, sge, num_sge);
if (rc) {
- pr_dbg("Error: Failed to build host SGE array\n");
complete_work(IBV_WC_GENERAL_ERR, rc, ctx);
goto out_dealloc_cqe_ctx;
}
@@ -609,10 +578,9 @@ void rdma_backend_post_recv(RdmaBackendDev *backend_dev,
wr.sg_list = new_sge;
wr.wr_id = bctx_id;
rc = ibv_post_recv(qp->ibqp, &wr, &bad_wr);
- pr_dbg("ibv_post_recv=%d\n", rc);
if (rc) {
- pr_dbg("Fail (%d, %d) to post recv WQE to qpn %d\n", rc, errno,
- qp->ibqp->qp_num);
+ rdma_error_report("ibv_post_recv fail, qpn=0x%x, rc=%d, errno=%d",
+ qp->ibqp->qp_num, rc, errno);
complete_work(IBV_WC_GENERAL_ERR, VENDOR_ERR_FAIL_BACKEND, ctx);
goto out_dealloc_cqe_ctx;
}
@@ -630,7 +598,12 @@ int rdma_backend_create_pd(RdmaBackendDev *backend_dev, RdmaBackendPD *pd)
{
pd->ibpd = ibv_alloc_pd(backend_dev->context);
- return pd->ibpd ? 0 : -EIO;
+ if (!pd->ibpd) {
+ rdma_error_report("ibv_alloc_pd fail, errno=%d", errno);
+ return -EIO;
+ }
+
+ return 0;
}
void rdma_backend_destroy_pd(RdmaBackendPD *pd)
@@ -643,16 +616,15 @@ void rdma_backend_destroy_pd(RdmaBackendPD *pd)
int rdma_backend_create_mr(RdmaBackendMR *mr, RdmaBackendPD *pd, void *addr,
size_t length, int access)
{
- pr_dbg("addr=0x%p\n", addr);
- pr_dbg("len=%zu\n", length);
mr->ibmr = ibv_reg_mr(pd->ibpd, addr, length, access);
- if (mr->ibmr) {
- pr_dbg("lkey=0x%x\n", mr->ibmr->lkey);
- pr_dbg("rkey=0x%x\n", mr->ibmr->rkey);
- mr->ibpd = pd->ibpd;
+ if (!mr->ibmr) {
+ rdma_error_report("ibv_reg_mr fail, errno=%d", errno);
+ return -EIO;
}
- return mr->ibmr ? 0 : -EIO;
+ mr->ibpd = pd->ibpd;
+
+ return 0;
}
void rdma_backend_destroy_mr(RdmaBackendMR *mr)
@@ -667,21 +639,21 @@ int rdma_backend_create_cq(RdmaBackendDev *backend_dev, RdmaBackendCQ *cq,
{
int rc;
- pr_dbg("cqe=%d\n", cqe);
-
- pr_dbg("dev->channel=%p\n", backend_dev->channel);
cq->ibcq = ibv_create_cq(backend_dev->context, cqe + 1, NULL,
backend_dev->channel, 0);
+ if (!cq->ibcq) {
+ rdma_error_report("ibv_create_cq fail, errno=%d", errno);
+ return -EIO;
+ }
- if (cq->ibcq) {
- rc = ibv_req_notify_cq(cq->ibcq, 0);
- if (rc) {
- pr_dbg("Error %d from ibv_req_notify_cq\n", rc);
- }
- cq->backend_dev = backend_dev;
+ rc = ibv_req_notify_cq(cq->ibcq, 0);
+ if (rc) {
+ rdma_warn_report("ibv_req_notify_cq fail, rc=%d, errno=%d", rc, errno);
}
- return cq->ibcq ? 0 : -EIO;
+ cq->backend_dev = backend_dev;
+
+ return 0;
}
void rdma_backend_destroy_cq(RdmaBackendCQ *cq)
@@ -700,7 +672,6 @@ int rdma_backend_create_qp(RdmaBackendQP *qp, uint8_t qp_type,
struct ibv_qp_init_attr attr = {0};
qp->ibqp = 0;
- pr_dbg("qp_type=%d\n", qp_type);
switch (qp_type) {
case IBV_QPT_GSI:
@@ -713,7 +684,7 @@ int rdma_backend_create_qp(RdmaBackendQP *qp, uint8_t qp_type,
break;
default:
- pr_dbg("Unsupported QP type %d\n", qp_type);
+ rdma_error_report("Unsupported QP type %d", qp_type);
return -EIO;
}
@@ -725,14 +696,9 @@ int rdma_backend_create_qp(RdmaBackendQP *qp, uint8_t qp_type,
attr.cap.max_send_sge = max_send_sge;
attr.cap.max_recv_sge = max_recv_sge;
- pr_dbg("max_send_wr=%d\n", max_send_wr);
- pr_dbg("max_recv_wr=%d\n", max_recv_wr);
- pr_dbg("max_send_sge=%d\n", max_send_sge);
- pr_dbg("max_recv_sge=%d\n", max_recv_sge);
-
qp->ibqp = ibv_create_qp(pd->ibpd, &attr);
- if (likely(!qp->ibqp)) {
- pr_dbg("Error from ibv_create_qp\n");
+ if (!qp->ibqp) {
+ rdma_error_report("ibv_create_qp fail, errno=%d", errno);
return -EIO;
}
@@ -740,8 +706,6 @@ int rdma_backend_create_qp(RdmaBackendQP *qp, uint8_t qp_type,
/* TODO: Query QP to get max_inline_data and save it to be used in send */
- pr_dbg("qpn=0x%x\n", qp->ibqp->qp_num);
-
return 0;
}
@@ -751,9 +715,6 @@ int rdma_backend_qp_state_init(RdmaBackendDev *backend_dev, RdmaBackendQP *qp,
struct ibv_qp_attr attr = {0};
int rc, attr_mask;
- pr_dbg("qpn=0x%x\n", qp->ibqp->qp_num);
- pr_dbg("sport_num=%d\n", backend_dev->port_num);
-
attr_mask = IBV_QP_STATE | IBV_QP_PKEY_INDEX | IBV_QP_PORT;
attr.qp_state = IBV_QPS_INIT;
attr.pkey_index = 0;
@@ -762,21 +723,23 @@ int rdma_backend_qp_state_init(RdmaBackendDev *backend_dev, RdmaBackendQP *qp,
switch (qp_type) {
case IBV_QPT_RC:
attr_mask |= IBV_QP_ACCESS_FLAGS;
+ trace_rdma_backend_rc_qp_state_init(qp->ibqp->qp_num);
break;
case IBV_QPT_UD:
attr.qkey = qkey;
attr_mask |= IBV_QP_QKEY;
+ trace_rdma_backend_ud_qp_state_init(qp->ibqp->qp_num, qkey);
break;
default:
- pr_dbg("Unsupported QP type %d\n", qp_type);
+ rdma_error_report("Unsupported QP type %d", qp_type);
return -EIO;
}
rc = ibv_modify_qp(qp->ibqp, &attr, attr_mask);
if (rc) {
- pr_dbg("Error %d from ibv_modify_qp\n", rc);
+ rdma_error_report("ibv_modify_qp fail, rc=%d, errno=%d", rc, errno);
return -EIO;
}
@@ -802,14 +765,6 @@ int rdma_backend_qp_state_rtr(RdmaBackendDev *backend_dev, RdmaBackendQP *qp,
switch (qp_type) {
case IBV_QPT_RC:
- pr_dbg("dgid=0x%" PRIx64 ",%" PRIx64 "\n",
- be64_to_cpu(ibv_gid.global.subnet_prefix),
- be64_to_cpu(ibv_gid.global.interface_id));
- pr_dbg("dqpn=0x%x\n", dqpn);
- pr_dbg("sgid_idx=%d\n", qp->sgid_idx);
- pr_dbg("sport_num=%d\n", backend_dev->port_num);
- pr_dbg("rq_psn=0x%x\n", rq_psn);
-
attr.path_mtu = IBV_MTU_1024;
attr.dest_qp_num = dqpn;
attr.max_dest_rd_atomic = 1;
@@ -824,20 +779,28 @@ int rdma_backend_qp_state_rtr(RdmaBackendDev *backend_dev, RdmaBackendQP *qp,
attr_mask |= IBV_QP_AV | IBV_QP_PATH_MTU | IBV_QP_DEST_QPN |
IBV_QP_RQ_PSN | IBV_QP_MAX_DEST_RD_ATOMIC |
IBV_QP_MIN_RNR_TIMER;
+
+ trace_rdma_backend_rc_qp_state_rtr(qp->ibqp->qp_num,
+ be64_to_cpu(ibv_gid.global.
+ subnet_prefix),
+ be64_to_cpu(ibv_gid.global.
+ interface_id),
+ qp->sgid_idx, dqpn, rq_psn);
break;
case IBV_QPT_UD:
- pr_dbg("qkey=0x%x\n", qkey);
if (use_qkey) {
attr.qkey = qkey;
attr_mask |= IBV_QP_QKEY;
}
+ trace_rdma_backend_ud_qp_state_rtr(qp->ibqp->qp_num, use_qkey ? qkey :
+ 0);
break;
}
rc = ibv_modify_qp(qp->ibqp, &attr, attr_mask);
if (rc) {
- pr_dbg("Error %d from ibv_modify_qp\n", rc);
+ rdma_error_report("ibv_modify_qp fail, rc=%d, errno=%d", rc, errno);
return -EIO;
}
@@ -850,9 +813,6 @@ int rdma_backend_qp_state_rts(RdmaBackendQP *qp, uint8_t qp_type,
struct ibv_qp_attr attr = {0};
int rc, attr_mask;
- pr_dbg("qpn=0x%x\n", qp->ibqp->qp_num);
- pr_dbg("sq_psn=0x%x\n", sq_psn);
-
attr.qp_state = IBV_QPS_RTS;
attr.sq_psn = sq_psn;
attr_mask = IBV_QP_STATE | IBV_QP_SQ_PSN;
@@ -866,20 +826,22 @@ int rdma_backend_qp_state_rts(RdmaBackendQP *qp, uint8_t qp_type,
attr_mask |= IBV_QP_TIMEOUT | IBV_QP_RETRY_CNT | IBV_QP_RNR_RETRY |
IBV_QP_MAX_QP_RD_ATOMIC;
+ trace_rdma_backend_rc_qp_state_rts(qp->ibqp->qp_num, sq_psn);
break;
case IBV_QPT_UD:
if (use_qkey) {
- pr_dbg("qkey=0x%x\n", qkey);
attr.qkey = qkey;
attr_mask |= IBV_QP_QKEY;
}
+ trace_rdma_backend_ud_qp_state_rts(qp->ibqp->qp_num, sq_psn,
+ use_qkey ? qkey : 0);
break;
}
rc = ibv_modify_qp(qp->ibqp, &attr, attr_mask);
if (rc) {
- pr_dbg("Error %d from ibv_modify_qp\n", rc);
+ rdma_error_report("ibv_modify_qp fail, rc=%d, errno=%d", rc, errno);
return -EIO;
}
@@ -890,7 +852,6 @@ int rdma_backend_query_qp(RdmaBackendQP *qp, struct ibv_qp_attr *attr,
int attr_mask, struct ibv_qp_init_attr *init_attr)
{
if (!qp->ibqp) {
- pr_dbg("QP1\n");
attr->qp_state = IBV_QPS_RTS;
return 0;
}
@@ -906,20 +867,23 @@ void rdma_backend_destroy_qp(RdmaBackendQP *qp)
}
#define CHK_ATTR(req, dev, member, fmt) ({ \
- pr_dbg("%s="fmt","fmt"\n", #member, dev.member, req->member); \
+ trace_rdma_check_dev_attr(#member, dev.member, req->member); \
if (req->member > dev.member) { \
- warn_report("%s = "fmt" is higher than host device capability "fmt, \
- #member, req->member, dev.member); \
+ rdma_warn_report("%s = "fmt" is higher than host device capability "fmt, \
+ #member, req->member, dev.member); \
req->member = dev.member; \
} \
- pr_dbg("%s="fmt"\n", #member, req->member); })
+})
static int init_device_caps(RdmaBackendDev *backend_dev,
struct ibv_device_attr *dev_attr)
{
struct ibv_device_attr bk_dev_attr;
+ int rc;
- if (ibv_query_device(backend_dev->context, &bk_dev_attr)) {
+ rc = ibv_query_device(backend_dev->context, &bk_dev_attr);
+ if (rc) {
+ rdma_error_report("ibv_query_device fail, rc=%d, errno=%d", rc, errno);
return -EIO;
}
@@ -928,9 +892,7 @@ static int init_device_caps(RdmaBackendDev *backend_dev,
CHK_ATTR(dev_attr, bk_dev_attr, max_mr_size, "%" PRId64);
CHK_ATTR(dev_attr, bk_dev_attr, max_qp, "%d");
CHK_ATTR(dev_attr, bk_dev_attr, max_sge, "%d");
- CHK_ATTR(dev_attr, bk_dev_attr, max_qp_wr, "%d");
CHK_ATTR(dev_attr, bk_dev_attr, max_cq, "%d");
- CHK_ATTR(dev_attr, bk_dev_attr, max_cqe, "%d");
CHK_ATTR(dev_attr, bk_dev_attr, max_mr, "%d");
CHK_ATTR(dev_attr, bk_dev_attr, max_pd, "%d");
CHK_ATTR(dev_attr, bk_dev_attr, max_qp_rd_atom, "%d");
@@ -946,10 +908,6 @@ static inline void build_mad_hdr(struct ibv_grh *grh, union ibv_gid *sgid,
grh->paylen = htons(paylen);
grh->sgid = *sgid;
grh->dgid = *my_gid;
-
- pr_dbg("paylen=%d (net=0x%x)\n", paylen, grh->paylen);
- pr_dbg("dgid=0x%llx\n", my_gid->global.interface_id);
- pr_dbg("sgid=0x%llx\n", sgid->global.interface_id);
}
static void process_incoming_mad_req(RdmaBackendDev *backend_dev,
@@ -960,21 +918,13 @@ static void process_incoming_mad_req(RdmaBackendDev *backend_dev,
BackendCtx *bctx;
char *mad;
- pr_dbg("umad_len=%d\n", msg->umad_len);
-
-#ifdef PVRDMA_DEBUG
- struct umad_hdr *hdr = (struct umad_hdr *)&msg->umad.mad;
- pr_dbg("bv %x cls %x cv %x mtd %x st %d tid %" PRIx64 " at %x atm %x\n",
- hdr->base_version, hdr->mgmt_class, hdr->class_version,
- hdr->method, hdr->status, be64toh(hdr->tid),
- hdr->attr_id, hdr->attr_mod);
-#endif
+ trace_mad_message("recv", msg->umad.mad, msg->umad_len);
qemu_mutex_lock(&backend_dev->recv_mads_list.lock);
o_ctx_id = qlist_pop(backend_dev->recv_mads_list.list);
qemu_mutex_unlock(&backend_dev->recv_mads_list.lock);
if (!o_ctx_id) {
- pr_dbg("No more free MADs buffers, waiting for a while\n");
+ rdma_warn_report("No more free MADs buffers, waiting for a while");
sleep(THR_POLL_TO);
return;
}
@@ -982,12 +932,10 @@ static void process_incoming_mad_req(RdmaBackendDev *backend_dev,
cqe_ctx_id = qnum_get_uint(qobject_to(QNum, o_ctx_id));
bctx = rdma_rm_get_cqe_ctx(backend_dev->rdma_dev_res, cqe_ctx_id);
if (unlikely(!bctx)) {
- pr_dbg("Error: Fail to find ctx for %ld\n", cqe_ctx_id);
+ rdma_error_report("No matching ctx for req %ld", cqe_ctx_id);
return;
}
- pr_dbg("id %ld, bctx %p, ctx %p\n", cqe_ctx_id, bctx, bctx->up_ctx);
-
mad = rdma_pci_dma_map(backend_dev->dev, bctx->sge.addr,
bctx->sge.length);
if (!mad || bctx->sge.length < msg->umad_len + MAD_HDR_SIZE) {
@@ -995,7 +943,6 @@ static void process_incoming_mad_req(RdmaBackendDev *backend_dev,
bctx->up_ctx);
} else {
struct ibv_wc wc = {0};
- pr_dbg_buf("mad", msg->umad.mad, msg->umad_len);
memset(mad, 0, bctx->sge.length);
build_mad_hdr((struct ibv_grh *)mad,
(union ibv_gid *)&msg->umad.hdr.addr.gid, &msg->hdr.sgid,
@@ -1025,13 +972,11 @@ static void rdmacm_mux_read(void *opaque, const uint8_t *buf, int size)
RdmaBackendDev *backend_dev = (RdmaBackendDev *)opaque;
RdmaCmMuxMsg *msg = (RdmaCmMuxMsg *)buf;
- pr_dbg("Got %d bytes\n", size);
- pr_dbg("msg_type=%d\n", msg->hdr.msg_type);
- pr_dbg("op_code=%d\n", msg->hdr.op_code);
+ trace_rdmacm_mux("read", msg->hdr.msg_type, msg->hdr.op_code);
if (msg->hdr.msg_type != RDMACM_MUX_MSG_TYPE_REQ &&
msg->hdr.op_code != RDMACM_MUX_OP_CODE_MAD) {
- pr_dbg("Error: Not a MAD request, skipping\n");
+ rdma_error_report("Error: Not a MAD request, skipping");
return;
}
process_incoming_mad_req(backend_dev, msg);
@@ -1045,7 +990,7 @@ static int mad_init(RdmaBackendDev *backend_dev, CharBackend *mad_chr_be)
ret = qemu_chr_fe_backend_connected(backend_dev->rdmacm_mux.chr_be);
if (!ret) {
- pr_dbg("Missing chardev for MAD multiplexer\n");
+ rdma_error_report("Missing chardev for MAD multiplexer");
return -EIO;
}
@@ -1063,7 +1008,6 @@ static int mad_init(RdmaBackendDev *backend_dev, CharBackend *mad_chr_be)
static void mad_fini(RdmaBackendDev *backend_dev)
{
- pr_dbg("Stopping MAD\n");
disable_rdmacm_mux_async(backend_dev);
qemu_chr_fe_disconnect(backend_dev->rdmacm_mux.chr_be);
if (backend_dev->recv_mads_list.list) {
@@ -1079,17 +1023,15 @@ int rdma_backend_get_gid_index(RdmaBackendDev *backend_dev,
int ret;
int i = 0;
- pr_dbg("0x%llx, 0x%llx\n",
- (long long unsigned int)be64_to_cpu(gid->global.subnet_prefix),
- (long long unsigned int)be64_to_cpu(gid->global.interface_id));
-
do {
ret = ibv_query_gid(backend_dev->context, backend_dev->port_num, i,
&sgid);
i++;
} while (!ret && (memcmp(&sgid, gid, sizeof(*gid))));
- pr_dbg("gid_index=%d\n", i - 1);
+ trace_rdma_backend_get_gid_index(be64_to_cpu(gid->global.subnet_prefix),
+ be64_to_cpu(gid->global.interface_id),
+ i - 1);
return ret ? ret : i - 1;
}
@@ -1100,16 +1042,15 @@ int rdma_backend_add_gid(RdmaBackendDev *backend_dev, const char *ifname,
RdmaCmMuxMsg msg = {};
int ret;
- pr_dbg("0x%llx, 0x%llx\n",
- (long long unsigned int)be64_to_cpu(gid->global.subnet_prefix),
- (long long unsigned int)be64_to_cpu(gid->global.interface_id));
+ trace_rdma_backend_gid_change("add", be64_to_cpu(gid->global.subnet_prefix),
+ be64_to_cpu(gid->global.interface_id));
msg.hdr.op_code = RDMACM_MUX_OP_CODE_REG;
memcpy(msg.hdr.sgid.raw, gid->raw, sizeof(msg.hdr.sgid));
- ret = exec_rdmacm_mux_req(backend_dev, &msg);
+ ret = rdmacm_mux_send(backend_dev, &msg);
if (ret) {
- pr_dbg("Fail to register GID to rdma_umadmux (%d)\n", ret);
+ rdma_error_report("Failed to register GID to rdma_umadmux (%d)", ret);
return -EIO;
}
@@ -1126,16 +1067,16 @@ int rdma_backend_del_gid(RdmaBackendDev *backend_dev, const char *ifname,
RdmaCmMuxMsg msg = {};
int ret;
- pr_dbg("0x%llx, 0x%llx\n",
- (long long unsigned int)be64_to_cpu(gid->global.subnet_prefix),
- (long long unsigned int)be64_to_cpu(gid->global.interface_id));
+ trace_rdma_backend_gid_change("del", be64_to_cpu(gid->global.subnet_prefix),
+ be64_to_cpu(gid->global.interface_id));
msg.hdr.op_code = RDMACM_MUX_OP_CODE_UNREG;
memcpy(msg.hdr.sgid.raw, gid->raw, sizeof(msg.hdr.sgid));
- ret = exec_rdmacm_mux_req(backend_dev, &msg);
+ ret = rdmacm_mux_send(backend_dev, &msg);
if (ret) {
- pr_dbg("Fail to unregister GID from rdma_umadmux (%d)\n", ret);
+ rdma_error_report("Failed to unregister GID from rdma_umadmux (%d)",
+ ret);
return -EIO;
}
@@ -1196,9 +1137,7 @@ int rdma_backend_init(RdmaBackendDev *backend_dev, PCIDevice *pdev,
backend_dev->ib_dev = *dev_list;
}
- pr_dbg("Using backend device %s, port %d\n",
- ibv_get_device_name(backend_dev->ib_dev), backend_dev->port_num);
- pr_dbg("uverb device %s\n", backend_dev->ib_dev->dev_name);
+ rdma_info_report("uverb device %s", backend_dev->ib_dev->dev_name);
backend_dev->context = ibv_open_device(backend_dev->ib_dev);
if (!backend_dev->context) {
@@ -1213,7 +1152,6 @@ int rdma_backend_init(RdmaBackendDev *backend_dev, PCIDevice *pdev,
ret = -EIO;
goto out_close_device;
}
- pr_dbg("dev->backend_dev.channel=%p\n", backend_dev->channel);
ret = init_device_caps(backend_dev, dev_attr);
if (ret) {
@@ -1225,7 +1163,7 @@ int rdma_backend_init(RdmaBackendDev *backend_dev, PCIDevice *pdev,
ret = mad_init(backend_dev, mad_chr_be);
if (ret) {
- error_setg(errp, "Fail to initialize mad");
+ error_setg(errp, "Failed to initialize mad");
ret = -EIO;
goto out_destroy_comm_channel;
}
@@ -1253,13 +1191,11 @@ out:
void rdma_backend_start(RdmaBackendDev *backend_dev)
{
- pr_dbg("Starting rdma_backend\n");
start_comp_thread(backend_dev);
}
void rdma_backend_stop(RdmaBackendDev *backend_dev)
{
- pr_dbg("Stopping rdma_backend\n");
stop_backend_thread(&backend_dev->comp_thread);
}
diff --git a/hw/rdma/rdma_rm.c b/hw/rdma/rdma_rm.c
index 268ff633a4..5dab4a2189 100644
--- a/hw/rdma/rdma_rm.c
+++ b/hw/rdma/rdma_rm.c
@@ -17,6 +17,7 @@
#include "qapi/error.h"
#include "cpu.h"
+#include "trace.h"
#include "rdma_utils.h"
#include "rdma_backend.h"
#include "rdma_rm.h"
@@ -49,25 +50,26 @@ static inline void res_tbl_free(RdmaRmResTbl *tbl)
g_free(tbl->bitmap);
}
-static inline void *res_tbl_get(RdmaRmResTbl *tbl, uint32_t handle)
+static inline void *rdma_res_tbl_get(RdmaRmResTbl *tbl, uint32_t handle)
{
- pr_dbg("%s, handle=%d\n", tbl->name, handle);
+ trace_rdma_res_tbl_get(tbl->name, handle);
if ((handle < tbl->tbl_sz) && (test_bit(handle, tbl->bitmap))) {
return tbl->tbl + handle * tbl->res_sz;
} else {
- pr_dbg("Invalid handle %d\n", handle);
+ rdma_error_report("Table %s, invalid handle %d", tbl->name, handle);
return NULL;
}
}
-static inline void *res_tbl_alloc(RdmaRmResTbl *tbl, uint32_t *handle)
+static inline void *rdma_res_tbl_alloc(RdmaRmResTbl *tbl, uint32_t *handle)
{
qemu_mutex_lock(&tbl->lock);
*handle = find_first_zero_bit(tbl->bitmap, tbl->tbl_sz);
if (*handle > tbl->tbl_sz) {
- pr_dbg("Failed to alloc, bitmap is full\n");
+ rdma_error_report("Table %s, failed to allocate, bitmap is full",
+ tbl->name);
qemu_mutex_unlock(&tbl->lock);
return NULL;
}
@@ -78,14 +80,14 @@ static inline void *res_tbl_alloc(RdmaRmResTbl *tbl, uint32_t *handle)
memset(tbl->tbl + *handle * tbl->res_sz, 0, tbl->res_sz);
- pr_dbg("%s, handle=%d\n", tbl->name, *handle);
+ trace_rdma_res_tbl_alloc(tbl->name, *handle);
return tbl->tbl + *handle * tbl->res_sz;
}
-static inline void res_tbl_dealloc(RdmaRmResTbl *tbl, uint32_t handle)
+static inline void rdma_res_tbl_dealloc(RdmaRmResTbl *tbl, uint32_t handle)
{
- pr_dbg("%s, handle=%d\n", tbl->name, handle);
+ trace_rdma_res_tbl_dealloc(tbl->name, handle);
qemu_mutex_lock(&tbl->lock);
@@ -102,7 +104,7 @@ int rdma_rm_alloc_pd(RdmaDeviceResources *dev_res, RdmaBackendDev *backend_dev,
RdmaRmPD *pd;
int ret = -ENOMEM;
- pd = res_tbl_alloc(&dev_res->pd_tbl, pd_handle);
+ pd = rdma_res_tbl_alloc(&dev_res->pd_tbl, pd_handle);
if (!pd) {
goto out;
}
@@ -118,7 +120,7 @@ int rdma_rm_alloc_pd(RdmaDeviceResources *dev_res, RdmaBackendDev *backend_dev,
return 0;
out_tbl_dealloc:
- res_tbl_dealloc(&dev_res->pd_tbl, *pd_handle);
+ rdma_res_tbl_dealloc(&dev_res->pd_tbl, *pd_handle);
out:
return ret;
@@ -126,7 +128,7 @@ out:
RdmaRmPD *rdma_rm_get_pd(RdmaDeviceResources *dev_res, uint32_t pd_handle)
{
- return res_tbl_get(&dev_res->pd_tbl, pd_handle);
+ return rdma_res_tbl_get(&dev_res->pd_tbl, pd_handle);
}
void rdma_rm_dealloc_pd(RdmaDeviceResources *dev_res, uint32_t pd_handle)
@@ -135,14 +137,14 @@ void rdma_rm_dealloc_pd(RdmaDeviceResources *dev_res, uint32_t pd_handle)
if (pd) {
rdma_backend_destroy_pd(&pd->backend_pd);
- res_tbl_dealloc(&dev_res->pd_tbl, pd_handle);
+ rdma_res_tbl_dealloc(&dev_res->pd_tbl, pd_handle);
}
}
int rdma_rm_alloc_mr(RdmaDeviceResources *dev_res, uint32_t pd_handle,
- uint64_t guest_start, size_t guest_length, void *host_virt,
- int access_flags, uint32_t *mr_handle, uint32_t *lkey,
- uint32_t *rkey)
+ uint64_t guest_start, uint64_t guest_length,
+ void *host_virt, int access_flags, uint32_t *mr_handle,
+ uint32_t *lkey, uint32_t *rkey)
{
RdmaRmMR *mr;
int ret = 0;
@@ -150,20 +152,15 @@ int rdma_rm_alloc_mr(RdmaDeviceResources *dev_res, uint32_t pd_handle,
pd = rdma_rm_get_pd(dev_res, pd_handle);
if (!pd) {
- pr_dbg("Invalid PD\n");
return -EINVAL;
}
- mr = res_tbl_alloc(&dev_res->mr_tbl, mr_handle);
+ mr = rdma_res_tbl_alloc(&dev_res->mr_tbl, mr_handle);
if (!mr) {
- pr_dbg("Failed to allocate obj in table\n");
return -ENOMEM;
}
- pr_dbg("mr_handle=%d\n", *mr_handle);
-
- pr_dbg("host_virt=0x%p\n", host_virt);
- pr_dbg("guest_start=0x%" PRIx64 "\n", guest_start);
- pr_dbg("length=%zu\n", guest_length);
+ trace_rdma_rm_alloc_mr(*mr_handle, host_virt, guest_start, guest_length,
+ access_flags);
if (host_virt) {
mr->virt = host_virt;
@@ -174,7 +171,6 @@ int rdma_rm_alloc_mr(RdmaDeviceResources *dev_res, uint32_t pd_handle,
ret = rdma_backend_create_mr(&mr->backend_mr, &pd->backend_pd, mr->virt,
mr->length, access_flags);
if (ret) {
- pr_dbg("Fail in rdma_backend_create_mr, err=%d\n", ret);
ret = -EIO;
goto out_dealloc_mr;
}
@@ -189,14 +185,14 @@ int rdma_rm_alloc_mr(RdmaDeviceResources *dev_res, uint32_t pd_handle,
return 0;
out_dealloc_mr:
- res_tbl_dealloc(&dev_res->mr_tbl, *mr_handle);
+ rdma_res_tbl_dealloc(&dev_res->mr_tbl, *mr_handle);
return ret;
}
RdmaRmMR *rdma_rm_get_mr(RdmaDeviceResources *dev_res, uint32_t mr_handle)
{
- return res_tbl_get(&dev_res->mr_tbl, mr_handle);
+ return rdma_res_tbl_get(&dev_res->mr_tbl, mr_handle);
}
void rdma_rm_dealloc_mr(RdmaDeviceResources *dev_res, uint32_t mr_handle)
@@ -205,12 +201,12 @@ void rdma_rm_dealloc_mr(RdmaDeviceResources *dev_res, uint32_t mr_handle)
if (mr) {
rdma_backend_destroy_mr(&mr->backend_mr);
- pr_dbg("start=0x%" PRIx64 "\n", mr->start);
+ trace_rdma_rm_dealloc_mr(mr_handle, mr->start);
if (mr->start) {
mr->virt -= (mr->start & (TARGET_PAGE_SIZE - 1));
munmap(mr->virt, mr->length);
}
- res_tbl_dealloc(&dev_res->mr_tbl, mr_handle);
+ rdma_res_tbl_dealloc(&dev_res->mr_tbl, mr_handle);
}
}
@@ -222,12 +218,13 @@ int rdma_rm_alloc_uc(RdmaDeviceResources *dev_res, uint32_t pfn,
/* TODO: Need to make sure pfn is between bar start address and
* bsd+RDMA_BAR2_UAR_SIZE
if (pfn > RDMA_BAR2_UAR_SIZE) {
- pr_err("pfn out of range (%d > %d)\n", pfn, RDMA_BAR2_UAR_SIZE);
+ rdma_error_report("pfn out of range (%d > %d)", pfn,
+ RDMA_BAR2_UAR_SIZE);
return -ENOMEM;
}
*/
- uc = res_tbl_alloc(&dev_res->uc_tbl, uc_handle);
+ uc = rdma_res_tbl_alloc(&dev_res->uc_tbl, uc_handle);
if (!uc) {
return -ENOMEM;
}
@@ -237,7 +234,7 @@ int rdma_rm_alloc_uc(RdmaDeviceResources *dev_res, uint32_t pfn,
RdmaRmUC *rdma_rm_get_uc(RdmaDeviceResources *dev_res, uint32_t uc_handle)
{
- return res_tbl_get(&dev_res->uc_tbl, uc_handle);
+ return rdma_res_tbl_get(&dev_res->uc_tbl, uc_handle);
}
void rdma_rm_dealloc_uc(RdmaDeviceResources *dev_res, uint32_t uc_handle)
@@ -245,13 +242,13 @@ void rdma_rm_dealloc_uc(RdmaDeviceResources *dev_res, uint32_t uc_handle)
RdmaRmUC *uc = rdma_rm_get_uc(dev_res, uc_handle);
if (uc) {
- res_tbl_dealloc(&dev_res->uc_tbl, uc_handle);
+ rdma_res_tbl_dealloc(&dev_res->uc_tbl, uc_handle);
}
}
RdmaRmCQ *rdma_rm_get_cq(RdmaDeviceResources *dev_res, uint32_t cq_handle)
{
- return res_tbl_get(&dev_res->cq_tbl, cq_handle);
+ return rdma_res_tbl_get(&dev_res->cq_tbl, cq_handle);
}
int rdma_rm_alloc_cq(RdmaDeviceResources *dev_res, RdmaBackendDev *backend_dev,
@@ -260,7 +257,7 @@ int rdma_rm_alloc_cq(RdmaDeviceResources *dev_res, RdmaBackendDev *backend_dev,
int rc;
RdmaRmCQ *cq;
- cq = res_tbl_alloc(&dev_res->cq_tbl, cq_handle);
+ cq = rdma_res_tbl_alloc(&dev_res->cq_tbl, cq_handle);
if (!cq) {
return -ENOMEM;
}
@@ -287,8 +284,6 @@ void rdma_rm_req_notify_cq(RdmaDeviceResources *dev_res, uint32_t cq_handle,
{
RdmaRmCQ *cq;
- pr_dbg("cq_handle=%d, notify=0x%x\n", cq_handle, notify);
-
cq = rdma_rm_get_cq(dev_res, cq_handle);
if (!cq) {
return;
@@ -297,8 +292,6 @@ void rdma_rm_req_notify_cq(RdmaDeviceResources *dev_res, uint32_t cq_handle,
if (cq->notify != CNT_SET) {
cq->notify = notify ? CNT_ARM : CNT_CLEAR;
}
-
- pr_dbg("notify=%d\n", cq->notify);
}
void rdma_rm_dealloc_cq(RdmaDeviceResources *dev_res, uint32_t cq_handle)
@@ -312,7 +305,7 @@ void rdma_rm_dealloc_cq(RdmaDeviceResources *dev_res, uint32_t cq_handle)
rdma_backend_destroy_cq(&cq->backend_cq);
- res_tbl_dealloc(&dev_res->cq_tbl, cq_handle);
+ rdma_res_tbl_dealloc(&dev_res->cq_tbl, cq_handle);
}
RdmaRmQP *rdma_rm_get_qp(RdmaDeviceResources *dev_res, uint32_t qpn)
@@ -323,6 +316,10 @@ RdmaRmQP *rdma_rm_get_qp(RdmaDeviceResources *dev_res, uint32_t qpn)
g_bytes_unref(key);
+ if (!qp) {
+ rdma_error_report("Invalid QP handle %d", qpn);
+ }
+
return qp;
}
@@ -338,11 +335,8 @@ int rdma_rm_alloc_qp(RdmaDeviceResources *dev_res, uint32_t pd_handle,
RdmaRmPD *pd;
uint32_t rm_qpn;
- pr_dbg("qp_type=%d\n", qp_type);
-
pd = rdma_rm_get_pd(dev_res, pd_handle);
if (!pd) {
- pr_err("Invalid pd handle (%d)\n", pd_handle);
return -EINVAL;
}
@@ -350,8 +344,8 @@ int rdma_rm_alloc_qp(RdmaDeviceResources *dev_res, uint32_t pd_handle,
rcq = rdma_rm_get_cq(dev_res, recv_cq_handle);
if (!scq || !rcq) {
- pr_err("Invalid send_cqn or recv_cqn (%d, %d)\n",
- send_cq_handle, recv_cq_handle);
+ rdma_error_report("Invalid send_cqn or recv_cqn (%d, %d)",
+ send_cq_handle, recv_cq_handle);
return -EINVAL;
}
@@ -360,11 +354,10 @@ int rdma_rm_alloc_qp(RdmaDeviceResources *dev_res, uint32_t pd_handle,
rcq->notify = CNT_SET;
}
- qp = res_tbl_alloc(&dev_res->qp_tbl, &rm_qpn);
+ qp = rdma_res_tbl_alloc(&dev_res->qp_tbl, &rm_qpn);
if (!qp) {
return -ENOMEM;
}
- pr_dbg("rm_qpn=%d\n", rm_qpn);
qp->qpn = rm_qpn;
qp->qp_state = IBV_QPS_RESET;
@@ -382,13 +375,13 @@ int rdma_rm_alloc_qp(RdmaDeviceResources *dev_res, uint32_t pd_handle,
}
*qpn = rdma_backend_qpn(&qp->backend_qp);
- pr_dbg("rm_qpn=%d, backend_qpn=0x%x\n", rm_qpn, *qpn);
+ trace_rdma_rm_alloc_qp(rm_qpn, *qpn, qp_type);
g_hash_table_insert(dev_res->qp_hash, g_bytes_new(qpn, sizeof(*qpn)), qp);
return 0;
out_dealloc_qp:
- res_tbl_dealloc(&dev_res->qp_tbl, qp->qpn);
+ rdma_res_tbl_dealloc(&dev_res->qp_tbl, qp->qpn);
return rc;
}
@@ -402,28 +395,22 @@ int rdma_rm_modify_qp(RdmaDeviceResources *dev_res, RdmaBackendDev *backend_dev,
RdmaRmQP *qp;
int ret;
- pr_dbg("qpn=0x%x\n", qp_handle);
- pr_dbg("qkey=0x%x\n", qkey);
-
qp = rdma_rm_get_qp(dev_res, qp_handle);
if (!qp) {
return -EINVAL;
}
- pr_dbg("qp_type=%d\n", qp->qp_type);
- pr_dbg("attr_mask=0x%x\n", attr_mask);
-
if (qp->qp_type == IBV_QPT_SMI) {
- pr_dbg("QP0 unsupported\n");
+ rdma_error_report("Got QP0 request");
return -EPERM;
} else if (qp->qp_type == IBV_QPT_GSI) {
- pr_dbg("QP1\n");
return 0;
}
+ trace_rdma_rm_modify_qp(qp_handle, attr_mask, qp_state, sgid_idx);
+
if (attr_mask & IBV_QP_STATE) {
qp->qp_state = qp_state;
- pr_dbg("qp_state=%d\n", qp->qp_state);
if (qp->qp_state == IBV_QPS_INIT) {
ret = rdma_backend_qp_state_init(backend_dev, &qp->backend_qp,
@@ -435,11 +422,11 @@ int rdma_rm_modify_qp(RdmaDeviceResources *dev_res, RdmaBackendDev *backend_dev,
if (qp->qp_state == IBV_QPS_RTR) {
/* Get backend gid index */
- pr_dbg("Guest sgid_idx=%d\n", sgid_idx);
sgid_idx = rdma_rm_get_backend_gid_index(dev_res, backend_dev,
sgid_idx);
if (sgid_idx <= 0) { /* TODO check also less than bk.max_sgid */
- pr_dbg("Fail to get bk sgid_idx for sgid_idx %d\n", sgid_idx);
+ rdma_error_report("Failed to get bk sgid_idx for sgid_idx %d",
+ sgid_idx);
return -EIO;
}
@@ -471,15 +458,11 @@ int rdma_rm_query_qp(RdmaDeviceResources *dev_res, RdmaBackendDev *backend_dev,
{
RdmaRmQP *qp;
- pr_dbg("qpn=0x%x\n", qp_handle);
-
qp = rdma_rm_get_qp(dev_res, qp_handle);
if (!qp) {
return -EINVAL;
}
- pr_dbg("qp_type=%d\n", qp->qp_type);
-
return rdma_backend_query_qp(&qp->backend_qp, attr, attr_mask, init_attr);
}
@@ -499,20 +482,18 @@ void rdma_rm_dealloc_qp(RdmaDeviceResources *dev_res, uint32_t qp_handle)
rdma_backend_destroy_qp(&qp->backend_qp);
- res_tbl_dealloc(&dev_res->qp_tbl, qp->qpn);
+ rdma_res_tbl_dealloc(&dev_res->qp_tbl, qp->qpn);
}
void *rdma_rm_get_cqe_ctx(RdmaDeviceResources *dev_res, uint32_t cqe_ctx_id)
{
void **cqe_ctx;
- cqe_ctx = res_tbl_get(&dev_res->cqe_ctx_tbl, cqe_ctx_id);
+ cqe_ctx = rdma_res_tbl_get(&dev_res->cqe_ctx_tbl, cqe_ctx_id);
if (!cqe_ctx) {
return NULL;
}
- pr_dbg("ctx=%p\n", *cqe_ctx);
-
return *cqe_ctx;
}
@@ -521,12 +502,11 @@ int rdma_rm_alloc_cqe_ctx(RdmaDeviceResources *dev_res, uint32_t *cqe_ctx_id,
{
void **cqe_ctx;
- cqe_ctx = res_tbl_alloc(&dev_res->cqe_ctx_tbl, cqe_ctx_id);
+ cqe_ctx = rdma_res_tbl_alloc(&dev_res->cqe_ctx_tbl, cqe_ctx_id);
if (!cqe_ctx) {
return -ENOMEM;
}
- pr_dbg("ctx=%p\n", ctx);
*cqe_ctx = ctx;
return 0;
@@ -534,7 +514,7 @@ int rdma_rm_alloc_cqe_ctx(RdmaDeviceResources *dev_res, uint32_t *cqe_ctx_id,
void rdma_rm_dealloc_cqe_ctx(RdmaDeviceResources *dev_res, uint32_t cqe_ctx_id)
{
- res_tbl_dealloc(&dev_res->cqe_ctx_tbl, cqe_ctx_id);
+ rdma_res_tbl_dealloc(&dev_res->cqe_ctx_tbl, cqe_ctx_id);
}
int rdma_rm_add_gid(RdmaDeviceResources *dev_res, RdmaBackendDev *backend_dev,
@@ -544,7 +524,6 @@ int rdma_rm_add_gid(RdmaDeviceResources *dev_res, RdmaBackendDev *backend_dev,
rc = rdma_backend_add_gid(backend_dev, ifname, gid);
if (rc) {
- pr_dbg("Fail to add gid\n");
return -EINVAL;
}
@@ -565,7 +544,6 @@ int rdma_rm_del_gid(RdmaDeviceResources *dev_res, RdmaBackendDev *backend_dev,
rc = rdma_backend_del_gid(backend_dev, ifname,
&dev_res->port.gid_tbl[gid_idx].gid);
if (rc) {
- pr_dbg("Fail to delete gid\n");
return -EINVAL;
}
@@ -580,7 +558,7 @@ int rdma_rm_get_backend_gid_index(RdmaDeviceResources *dev_res,
RdmaBackendDev *backend_dev, int sgid_idx)
{
if (unlikely(sgid_idx < 0 || sgid_idx >= MAX_PORT_GIDS)) {
- pr_dbg("Got invalid sgid_idx %d\n", sgid_idx);
+ rdma_error_report("Got invalid sgid_idx %d", sgid_idx);
return -EINVAL;
}
@@ -590,9 +568,6 @@ int rdma_rm_get_backend_gid_index(RdmaDeviceResources *dev_res,
&dev_res->port.gid_tbl[sgid_idx].gid);
}
- pr_dbg("backend_gid_index=%d\n",
- dev_res->port.gid_tbl[sgid_idx].backend_gid_index);
-
return dev_res->port.gid_tbl[sgid_idx].backend_gid_index;
}
diff --git a/hw/rdma/rdma_rm.h b/hw/rdma/rdma_rm.h
index 3c602c04c0..9ec87d2667 100644
--- a/hw/rdma/rdma_rm.h
+++ b/hw/rdma/rdma_rm.h
@@ -31,9 +31,9 @@ RdmaRmPD *rdma_rm_get_pd(RdmaDeviceResources *dev_res, uint32_t pd_handle);
void rdma_rm_dealloc_pd(RdmaDeviceResources *dev_res, uint32_t pd_handle);
int rdma_rm_alloc_mr(RdmaDeviceResources *dev_res, uint32_t pd_handle,
- uint64_t guest_start, size_t guest_length, void *host_virt,
- int access_flags, uint32_t *mr_handle, uint32_t *lkey,
- uint32_t *rkey);
+ uint64_t guest_start, uint64_t guest_length,
+ void *host_virt, int access_flags, uint32_t *mr_handle,
+ uint32_t *lkey, uint32_t *rkey);
RdmaRmMR *rdma_rm_get_mr(RdmaDeviceResources *dev_res, uint32_t mr_handle);
void rdma_rm_dealloc_mr(RdmaDeviceResources *dev_res, uint32_t mr_handle);
diff --git a/hw/rdma/rdma_utils.c b/hw/rdma/rdma_utils.c
index 4fbea8cde2..b9f07fcda7 100644
--- a/hw/rdma/rdma_utils.c
+++ b/hw/rdma/rdma_utils.c
@@ -14,26 +14,23 @@
*/
#include "qemu/osdep.h"
+#include "trace.h"
#include "rdma_utils.h"
-#ifdef PVRDMA_DEBUG
-unsigned long pr_dbg_cnt;
-#endif
-
void *rdma_pci_dma_map(PCIDevice *dev, dma_addr_t addr, dma_addr_t plen)
{
void *p;
hwaddr len = plen;
if (!addr) {
- pr_dbg("addr is NULL\n");
+ rdma_error_report("addr is NULL");
return NULL;
}
p = pci_dma_map(dev, addr, &len, DMA_DIRECTION_TO_DEVICE);
if (!p) {
- pr_dbg("Fail in pci_dma_map, addr=0x%" PRIx64 ", len=%" PRId64 "\n",
- addr, len);
+ rdma_error_report("pci_dma_map fail, addr=0x%"PRIx64", len=%"PRId64,
+ addr, len);
return NULL;
}
@@ -42,14 +39,14 @@ void *rdma_pci_dma_map(PCIDevice *dev, dma_addr_t addr, dma_addr_t plen)
return NULL;
}
- pr_dbg("0x%" PRIx64 " -> %p (len=% " PRId64 ")\n", addr, p, len);
+ trace_rdma_pci_dma_map(addr, p, len);
return p;
}
void rdma_pci_dma_unmap(PCIDevice *dev, void *buffer, dma_addr_t len)
{
- pr_dbg("%p\n", buffer);
+ trace_rdma_pci_dma_unmap(buffer);
if (buffer) {
pci_dma_unmap(dev, buffer, len, DMA_DIRECTION_TO_DEVICE, 0);
}
diff --git a/hw/rdma/rdma_utils.h b/hw/rdma/rdma_utils.h
index 4490ea0b94..acd148837f 100644
--- a/hw/rdma/rdma_utils.h
+++ b/hw/rdma/rdma_utils.h
@@ -17,48 +17,17 @@
#ifndef RDMA_UTILS_H
#define RDMA_UTILS_H
+#include "qemu/error-report.h"
#include "hw/pci/pci.h"
#include "sysemu/dma.h"
#include "stdio.h"
-#define pr_info(fmt, ...) \
- fprintf(stdout, "%s: %-20s (%3d): " fmt, "rdma", __func__, __LINE__,\
- ## __VA_ARGS__)
-
-#define pr_err(fmt, ...) \
- fprintf(stderr, "%s: Error at %-20s (%3d): " fmt, "rdma", __func__, \
- __LINE__, ## __VA_ARGS__)
-
-#ifdef PVRDMA_DEBUG
-extern unsigned long pr_dbg_cnt;
-
-#define init_pr_dbg(void) \
-{ \
- pr_dbg_cnt = 0; \
-}
-
-#define pr_dbg(fmt, ...) \
- fprintf(stdout, "%lx %ld: %-20s (%3d): " fmt, pthread_self(), pr_dbg_cnt++, \
- __func__, __LINE__, ## __VA_ARGS__)
-
-#define pr_dbg_buf(title, buf, len) \
-{ \
- int i; \
- char *b = g_malloc0(len * 3 + 1); \
- char b1[4]; \
- for (i = 0; i < len; i++) { \
- sprintf(b1, "%.2X ", buf[i] & 0x000000FF); \
- strcat(b, b1); \
- } \
- pr_dbg("%s (%d): %s\n", title, len, b); \
- g_free(b); \
-}
-
-#else
-#define init_pr_dbg(void)
-#define pr_dbg(fmt, ...)
-#define pr_dbg_buf(title, buf, len)
-#endif
+#define rdma_error_report(fmt, ...) \
+ error_report("%s: " fmt, "rdma", ## __VA_ARGS__)
+#define rdma_warn_report(fmt, ...) \
+ warn_report("%s: " fmt, "rdma", ## __VA_ARGS__)
+#define rdma_info_report(fmt, ...) \
+ info_report("%s: " fmt, "rdma", ## __VA_ARGS__)
void *rdma_pci_dma_map(PCIDevice *dev, dma_addr_t addr, dma_addr_t plen);
void rdma_pci_dma_unmap(PCIDevice *dev, void *buffer, dma_addr_t len);
diff --git a/hw/rdma/trace-events b/hw/rdma/trace-events
index c4c202e647..0fad56c882 100644
--- a/hw/rdma/trace-events
+++ b/hw/rdma/trace-events
@@ -1,5 +1,31 @@
# See docs/tracing.txt for syntax documentation.
-#hw/rdma/rdma_backend.c
-create_ah_cache_hit(uint64_t subnet, uint64_t net_id) "subnet = 0x%"PRIx64" net_id = 0x%"PRIx64
-create_ah_cache_miss(uint64_t subnet, uint64_t net_id) "subnet = 0x%"PRIx64" net_id = 0x%"PRIx64
+# hw/rdma/rdma_backend.c
+rdma_check_dev_attr(const char *name, int max_bk, int max_fe) "%s: be=%d, fe=%d"
+rdma_create_ah_cache_hit(uint64_t subnet, uint64_t if_id) "subnet=0x%"PRIx64",if_id=0x%"PRIx64
+rdma_create_ah_cache_miss(uint64_t subnet, uint64_t if_id) "subnet=0x%"PRIx64",if_id=0x%"PRIx64
+rdma_poll_cq(int ne, void *ibcq) "Got %d completion(s) from cq %p"
+rdmacm_mux(const char *title, int msg_type, int op_code) "%s: msg_type=%d, op_code=%d"
+rdmacm_mux_check_op_status(int msg_type, int op_code, int err_code) "resp: msg_type=%d, op_code=%d, err_code=%d"
+rdma_mad_message(const char *title, int len, char *data) "mad %s (%d): %s"
+rdma_backend_rc_qp_state_init(uint32_t qpn) "RC QP 0x%x switch to INIT"
+rdma_backend_ud_qp_state_init(uint32_t qpn, uint32_t qkey) "UD QP 0x%x switch to INIT, qkey=0x%x"
+rdma_backend_rc_qp_state_rtr(uint32_t qpn, uint64_t subnet, uint64_t ifid, uint8_t sgid_idx, uint32_t dqpn, uint32_t rq_psn) "RC QP 0x%x switch to RTR, subnet = 0x%"PRIx64", ifid = 0x%"PRIx64 ", sgid_idx=%d, dqpn=0x%x, rq_psn=0x%x"
+rdma_backend_ud_qp_state_rtr(uint32_t qpn, uint32_t qkey) "UD QP 0x%x switch to RTR, qkey=0x%x"
+rdma_backend_rc_qp_state_rts(uint32_t qpn, uint32_t sq_psn) "RC QP 0x%x switch to RTS, sq_psn=0x%x, "
+rdma_backend_ud_qp_state_rts(uint32_t qpn, uint32_t sq_psn, uint32_t qkey) "UD QP 0x%x switch to RTS, sq_psn=0x%x, qkey=0x%x"
+rdma_backend_get_gid_index(uint64_t subnet, uint64_t ifid, int gid_idx) "subnet=0x%"PRIx64", ifid=0x%"PRIx64 ", gid_idx=%d"
+rdma_backend_gid_change(const char *op, uint64_t subnet, uint64_t ifid) "%s subnet=0x%"PRIx64", ifid=0x%"PRIx64
+
+# hw/rdma/rdma_rm.c
+rdma_res_tbl_get(char *name, uint32_t handle) "tbl %s, handle %d"
+rdma_res_tbl_alloc(char *name, uint32_t handle) "tbl %s, handle %d"
+rdma_res_tbl_dealloc(char *name, uint32_t handle) "tbl %s, handle %d"
+rdma_rm_alloc_mr(uint32_t mr_handle, void *host_virt, uint64_t guest_start, uint64_t guest_length, int access_flags) "mr_handle=%d, host_virt=%p, guest_start=0x%"PRIx64", length=%" PRId64", access_flags=0x%x"
+rdma_rm_dealloc_mr(uint32_t mr_handle, uint64_t guest_start) "mr_handle=%d, guest_start=0x%"PRIx64
+rdma_rm_alloc_qp(uint32_t rm_qpn, uint32_t backend_qpn, uint8_t qp_type) "rm_qpn=%d, backend_qpn=0x%x, qp_type=%d"
+rdma_rm_modify_qp(uint32_t qpn, uint32_t attr_mask, int qp_state, uint8_t sgid_idx) "qpn=0x%x, attr_mask=0x%x, qp_state=%d, sgid_idx=%d"
+
+# hw/rdma/rdma_utils.c
+rdma_pci_dma_map(uint64_t addr, void *vaddr, uint64_t len) "0x%"PRIx64" -> %p (len=%" PRId64")"
+rdma_pci_dma_unmap(void *vaddr) "%p"
diff --git a/hw/rdma/vmw/pvrdma.h b/hw/rdma/vmw/pvrdma.h
index ffae36986e..0879224957 100644
--- a/hw/rdma/vmw/pvrdma.h
+++ b/hw/rdma/vmw/pvrdma.h
@@ -127,6 +127,6 @@ static inline void post_interrupt(PVRDMADev *dev, unsigned vector)
}
}
-int execute_command(PVRDMADev *dev);
+int pvrdma_exec_cmd(PVRDMADev *dev);
#endif
diff --git a/hw/rdma/vmw/pvrdma_cmd.c b/hw/rdma/vmw/pvrdma_cmd.c
index 89920887bf..21a55e225a 100644
--- a/hw/rdma/vmw/pvrdma_cmd.c
+++ b/hw/rdma/vmw/pvrdma_cmd.c
@@ -14,7 +14,6 @@
*/
#include "qemu/osdep.h"
-#include "qemu/error-report.h"
#include "cpu.h"
#include "hw/hw.h"
#include "hw/pci/pci.h"
@@ -24,6 +23,7 @@
#include "../rdma_rm.h"
#include "../rdma_utils.h"
+#include "trace.h"
#include "pvrdma.h"
#include "standard-headers/rdma/vmw_pvrdma-abi.h"
@@ -35,40 +35,38 @@ static void *pvrdma_map_to_pdir(PCIDevice *pdev, uint64_t pdir_dma,
void *host_virt = NULL, *curr_page;
if (!nchunks) {
- pr_dbg("nchunks=0\n");
+ rdma_error_report("Got nchunks=0");
return NULL;
}
dir = rdma_pci_dma_map(pdev, pdir_dma, TARGET_PAGE_SIZE);
if (!dir) {
- error_report("PVRDMA: Failed to map to page directory");
+ rdma_error_report("Failed to map to page directory");
return NULL;
}
tbl = rdma_pci_dma_map(pdev, dir[0], TARGET_PAGE_SIZE);
if (!tbl) {
- error_report("PVRDMA: Failed to map to page table 0");
+ rdma_error_report("Failed to map to page table 0");
goto out_unmap_dir;
}
curr_page = rdma_pci_dma_map(pdev, (dma_addr_t)tbl[0], TARGET_PAGE_SIZE);
if (!curr_page) {
- error_report("PVRDMA: Failed to map the first page");
+ rdma_error_report("Failed to map the page 0");
goto out_unmap_tbl;
}
host_virt = mremap(curr_page, 0, length, MREMAP_MAYMOVE);
- pr_dbg("mremap %p -> %p\n", curr_page, host_virt);
if (host_virt == MAP_FAILED) {
host_virt = NULL;
- error_report("PVRDMA: Failed to remap memory for host_virt");
+ rdma_error_report("Failed to remap memory for host_virt");
goto out_unmap_tbl;
}
+ trace_pvrdma_map_to_pdir_host_virt(curr_page, host_virt);
rdma_pci_dma_unmap(pdev, curr_page, TARGET_PAGE_SIZE);
- pr_dbg("host_virt=%p\n", host_virt);
-
dir_idx = 0;
tbl_idx = 1;
addr_idx = 1;
@@ -76,28 +74,28 @@ static void *pvrdma_map_to_pdir(PCIDevice *pdev, uint64_t pdir_dma,
if (tbl_idx == TARGET_PAGE_SIZE / sizeof(uint64_t)) {
tbl_idx = 0;
dir_idx++;
- pr_dbg("Mapping to table %d\n", dir_idx);
rdma_pci_dma_unmap(pdev, tbl, TARGET_PAGE_SIZE);
tbl = rdma_pci_dma_map(pdev, dir[dir_idx], TARGET_PAGE_SIZE);
if (!tbl) {
- error_report("PVRDMA: Failed to map to page table %d", dir_idx);
+ rdma_error_report("Failed to map to page table %d", dir_idx);
goto out_unmap_host_virt;
}
}
- pr_dbg("guest_dma[%d]=0x%" PRIx64 "\n", addr_idx, tbl[tbl_idx]);
-
curr_page = rdma_pci_dma_map(pdev, (dma_addr_t)tbl[tbl_idx],
TARGET_PAGE_SIZE);
if (!curr_page) {
- error_report("PVRDMA: Failed to map to page %d, dir %d", tbl_idx,
- dir_idx);
+ rdma_error_report("Failed to map to page %d, dir %d", tbl_idx,
+ dir_idx);
goto out_unmap_host_virt;
}
mremap(curr_page, 0, TARGET_PAGE_SIZE, MREMAP_MAYMOVE | MREMAP_FIXED,
host_virt + TARGET_PAGE_SIZE * addr_idx);
+ trace_pvrdma_map_to_pdir_next_page(addr_idx, curr_page, host_virt +
+ TARGET_PAGE_SIZE * addr_idx);
+
rdma_pci_dma_unmap(pdev, curr_page, TARGET_PAGE_SIZE);
addr_idx++;
@@ -127,7 +125,6 @@ static int query_port(PVRDMADev *dev, union pvrdma_cmd_req *req,
struct pvrdma_cmd_query_port_resp *resp = &rsp->query_port_resp;
struct pvrdma_port_attr attrs = {0};
- pr_dbg("port=%d\n", cmd->port_num);
if (cmd->port_num > MAX_PORTS) {
return -EINVAL;
}
@@ -159,12 +156,10 @@ static int query_pkey(PVRDMADev *dev, union pvrdma_cmd_req *req,
struct pvrdma_cmd_query_pkey *cmd = &req->query_pkey;
struct pvrdma_cmd_query_pkey_resp *resp = &rsp->query_pkey_resp;
- pr_dbg("port=%d\n", cmd->port_num);
if (cmd->port_num > MAX_PORTS) {
return -EINVAL;
}
- pr_dbg("index=%d\n", cmd->index);
if (cmd->index > MAX_PKEYS) {
return -EINVAL;
}
@@ -172,7 +167,6 @@ static int query_pkey(PVRDMADev *dev, union pvrdma_cmd_req *req,
memset(resp, 0, sizeof(*resp));
resp->pkey = PVRDMA_PKEY;
- pr_dbg("pkey=0x%x\n", resp->pkey);
return 0;
}
@@ -184,8 +178,6 @@ static int create_pd(PVRDMADev *dev, union pvrdma_cmd_req *req,
struct pvrdma_cmd_create_pd_resp *resp = &rsp->create_pd_resp;
int rc;
- pr_dbg("context=0x%x\n", cmd->ctx_handle ? cmd->ctx_handle : 0);
-
memset(resp, 0, sizeof(*resp));
rc = rdma_rm_alloc_pd(&dev->rdma_dev_res, &dev->backend_dev,
&resp->pd_handle, cmd->ctx_handle);
@@ -198,8 +190,6 @@ static int destroy_pd(PVRDMADev *dev, union pvrdma_cmd_req *req,
{
struct pvrdma_cmd_destroy_pd *cmd = &req->destroy_pd;
- pr_dbg("pd_handle=%d\n", cmd->pd_handle);
-
rdma_rm_dealloc_pd(&dev->rdma_dev_res, cmd->pd_handle);
return 0;
@@ -216,15 +206,11 @@ static int create_mr(PVRDMADev *dev, union pvrdma_cmd_req *req,
memset(resp, 0, sizeof(*resp));
- pr_dbg("pd_handle=%d\n", cmd->pd_handle);
- pr_dbg("access_flags=0x%x\n", cmd->access_flags);
- pr_dbg("flags=0x%x\n", cmd->flags);
-
if (!(cmd->flags & PVRDMA_MR_FLAG_DMA)) {
host_virt = pvrdma_map_to_pdir(pci_dev, cmd->pdir_dma, cmd->nchunks,
cmd->length);
if (!host_virt) {
- pr_dbg("Failed to map to pdir\n");
+ rdma_error_report("Failed to map to pdir");
return -EINVAL;
}
}
@@ -244,8 +230,6 @@ static int destroy_mr(PVRDMADev *dev, union pvrdma_cmd_req *req,
{
struct pvrdma_cmd_destroy_mr *cmd = &req->destroy_mr;
- pr_dbg("mr_handle=%d\n", cmd->mr_handle);
-
rdma_rm_dealloc_mr(&dev->rdma_dev_res, cmd->mr_handle);
return 0;
@@ -260,20 +244,19 @@ static int create_cq_ring(PCIDevice *pci_dev , PvrdmaRing **ring,
char ring_name[MAX_RING_NAME_SZ];
if (!nchunks || nchunks > PVRDMA_MAX_FAST_REG_PAGES) {
- pr_dbg("invalid nchunks: %d\n", nchunks);
+ rdma_error_report("Got invalid nchunks: %d", nchunks);
return rc;
}
- pr_dbg("pdir_dma=0x%llx\n", (long long unsigned int)pdir_dma);
dir = rdma_pci_dma_map(pci_dev, pdir_dma, TARGET_PAGE_SIZE);
if (!dir) {
- pr_dbg("Failed to map to CQ page directory\n");
+ rdma_error_report("Failed to map to CQ page directory");
goto out;
}
tbl = rdma_pci_dma_map(pci_dev, dir[0], TARGET_PAGE_SIZE);
if (!tbl) {
- pr_dbg("Failed to map to CQ page table\n");
+ rdma_error_report("Failed to map to CQ page table");
goto out;
}
@@ -284,7 +267,7 @@ static int create_cq_ring(PCIDevice *pci_dev , PvrdmaRing **ring,
rdma_pci_dma_map(pci_dev, tbl[0], TARGET_PAGE_SIZE);
if (!r->ring_state) {
- pr_dbg("Failed to map to CQ ring state\n");
+ rdma_error_report("Failed to map to CQ ring state");
goto out_free_ring;
}
@@ -339,8 +322,6 @@ static int create_cq(PVRDMADev *dev, union pvrdma_cmd_req *req,
return rc;
}
- pr_dbg("ring=%p\n", ring);
-
rc = rdma_rm_alloc_cq(&dev->rdma_dev_res, &dev->backend_dev, cmd->cqe,
&resp->cq_handle, ring);
if (rc) {
@@ -359,11 +340,9 @@ static int destroy_cq(PVRDMADev *dev, union pvrdma_cmd_req *req,
RdmaRmCQ *cq;
PvrdmaRing *ring;
- pr_dbg("cq_handle=%d\n", cmd->cq_handle);
-
cq = rdma_rm_get_cq(&dev->rdma_dev_res, cmd->cq_handle);
if (!cq) {
- pr_dbg("Invalid CQ handle\n");
+ rdma_error_report("Got invalid CQ handle");
return -EINVAL;
}
@@ -388,42 +367,33 @@ static int create_qp_rings(PCIDevice *pci_dev, uint64_t pdir_dma,
if (!spages || spages > PVRDMA_MAX_FAST_REG_PAGES
|| !rpages || rpages > PVRDMA_MAX_FAST_REG_PAGES) {
- pr_dbg("invalid pages: %d, %d\n", spages, rpages);
+ rdma_error_report("Got invalid page count for QP ring: %d, %d", spages,
+ rpages);
return rc;
}
- pr_dbg("pdir_dma=0x%llx\n", (long long unsigned int)pdir_dma);
dir = rdma_pci_dma_map(pci_dev, pdir_dma, TARGET_PAGE_SIZE);
if (!dir) {
- pr_dbg("Failed to map to CQ page directory\n");
+ rdma_error_report("Failed to map to CQ page directory");
goto out;
}
tbl = rdma_pci_dma_map(pci_dev, dir[0], TARGET_PAGE_SIZE);
if (!tbl) {
- pr_dbg("Failed to map to CQ page table\n");
+ rdma_error_report("Failed to map to CQ page table");
goto out;
}
sr = g_malloc(2 * sizeof(*rr));
rr = &sr[1];
- pr_dbg("sring=%p\n", sr);
- pr_dbg("rring=%p\n", rr);
*rings = sr;
- pr_dbg("scqe=%d\n", scqe);
- pr_dbg("smax_sge=%d\n", smax_sge);
- pr_dbg("spages=%d\n", spages);
- pr_dbg("rcqe=%d\n", rcqe);
- pr_dbg("rmax_sge=%d\n", rmax_sge);
- pr_dbg("rpages=%d\n", rpages);
-
/* Create send ring */
sr->ring_state = (struct pvrdma_ring *)
rdma_pci_dma_map(pci_dev, tbl[0], TARGET_PAGE_SIZE);
if (!sr->ring_state) {
- pr_dbg("Failed to map to CQ ring state\n");
+ rdma_error_report("Failed to map to CQ ring state");
goto out_free_sr_mem;
}
@@ -468,9 +438,7 @@ out:
static void destroy_qp_rings(PvrdmaRing *ring)
{
- pr_dbg("sring=%p\n", &ring[0]);
pvrdma_ring_free(&ring[0]);
- pr_dbg("rring=%p\n", &ring[1]);
pvrdma_ring_free(&ring[1]);
rdma_pci_dma_unmap(ring->dev, ring->ring_state, TARGET_PAGE_SIZE);
@@ -487,9 +455,6 @@ static int create_qp(PVRDMADev *dev, union pvrdma_cmd_req *req,
memset(resp, 0, sizeof(*resp));
- pr_dbg("total_chunks=%d\n", cmd->total_chunks);
- pr_dbg("send_chunks=%d\n", cmd->send_chunks);
-
rc = create_qp_rings(PCI_DEVICE(dev), cmd->pdir_dma, &rings,
cmd->max_send_wr, cmd->max_send_sge, cmd->send_chunks,
cmd->max_recv_wr, cmd->max_recv_sge,
@@ -498,8 +463,6 @@ static int create_qp(PVRDMADev *dev, union pvrdma_cmd_req *req,
return rc;
}
- pr_dbg("rings=%p\n", rings);
-
rc = rdma_rm_alloc_qp(&dev->rdma_dev_res, cmd->pd_handle, cmd->qp_type,
cmd->max_send_wr, cmd->max_send_sge,
cmd->send_cq_handle, cmd->max_recv_wr,
@@ -525,8 +488,6 @@ static int modify_qp(PVRDMADev *dev, union pvrdma_cmd_req *req,
struct pvrdma_cmd_modify_qp *cmd = &req->modify_qp;
int rc;
- pr_dbg("qp_handle=%d\n", cmd->qp_handle);
-
memset(rsp, 0, sizeof(*rsp));
/* No need to verify sgid_index since it is u8 */
@@ -551,9 +512,6 @@ static int query_qp(PVRDMADev *dev, union pvrdma_cmd_req *req,
struct ibv_qp_init_attr init_attr;
int rc;
- pr_dbg("qp_handle=%d\n", cmd->qp_handle);
- pr_dbg("attr_mask=0x%x\n", cmd->attr_mask);
-
memset(rsp, 0, sizeof(*rsp));
rc = rdma_rm_query_qp(&dev->rdma_dev_res, &dev->backend_dev, cmd->qp_handle,
@@ -572,7 +530,6 @@ static int destroy_qp(PVRDMADev *dev, union pvrdma_cmd_req *req,
qp = rdma_rm_get_qp(&dev->rdma_dev_res, cmd->qp_handle);
if (!qp) {
- pr_dbg("Invalid QP handle\n");
return -EINVAL;
}
@@ -591,16 +548,10 @@ static int create_bind(PVRDMADev *dev, union pvrdma_cmd_req *req,
int rc;
union ibv_gid *gid = (union ibv_gid *)&cmd->new_gid;
- pr_dbg("index=%d\n", cmd->index);
-
if (cmd->index >= MAX_PORT_GIDS) {
return -EINVAL;
}
- pr_dbg("gid[%d]=0x%llx,0x%llx\n", cmd->index,
- (long long unsigned int)be64_to_cpu(gid->global.subnet_prefix),
- (long long unsigned int)be64_to_cpu(gid->global.interface_id));
-
rc = rdma_rm_add_gid(&dev->rdma_dev_res, &dev->backend_dev,
dev->backend_eth_device_name, gid, cmd->index);
@@ -614,8 +565,6 @@ static int destroy_bind(PVRDMADev *dev, union pvrdma_cmd_req *req,
struct pvrdma_cmd_destroy_bind *cmd = &req->destroy_bind;
- pr_dbg("index=%d\n", cmd->index);
-
if (cmd->index >= MAX_PORT_GIDS) {
return -EINVAL;
}
@@ -633,8 +582,6 @@ static int create_uc(PVRDMADev *dev, union pvrdma_cmd_req *req,
struct pvrdma_cmd_create_uc_resp *resp = &rsp->create_uc_resp;
int rc;
- pr_dbg("pfn=%d\n", cmd->pfn);
-
memset(resp, 0, sizeof(*resp));
rc = rdma_rm_alloc_uc(&dev->rdma_dev_res, cmd->pfn, &resp->ctx_handle);
@@ -646,8 +593,6 @@ static int destroy_uc(PVRDMADev *dev, union pvrdma_cmd_req *req,
{
struct pvrdma_cmd_destroy_uc *cmd = &req->destroy_uc;
- pr_dbg("ctx_handle=%d\n", cmd->ctx_handle);
-
rdma_rm_dealloc_uc(&dev->rdma_dev_res, cmd->ctx_handle);
return 0;
@@ -680,22 +625,21 @@ static struct cmd_handler cmd_handlers[] = {
{PVRDMA_CMD_DESTROY_BIND, PVRDMA_CMD_DESTROY_BIND_RESP_NOOP, destroy_bind},
};
-int execute_command(PVRDMADev *dev)
+int pvrdma_exec_cmd(PVRDMADev *dev)
{
int err = 0xFFFF;
DSRInfo *dsr_info;
dsr_info = &dev->dsr_info;
- pr_dbg("cmd=%d\n", dsr_info->req->hdr.cmd);
if (dsr_info->req->hdr.cmd >= sizeof(cmd_handlers) /
sizeof(struct cmd_handler)) {
- pr_dbg("Unsupported command\n");
+ rdma_error_report("Unsupported command");
goto out;
}
if (!cmd_handlers[dsr_info->req->hdr.cmd].exec) {
- pr_dbg("Unsupported command (not implemented yet)\n");
+ rdma_error_report("Unsupported command (not implemented yet)");
goto out;
}
@@ -704,7 +648,8 @@ int execute_command(PVRDMADev *dev)
dsr_info->rsp->hdr.response = dsr_info->req->hdr.response;
dsr_info->rsp->hdr.ack = cmd_handlers[dsr_info->req->hdr.cmd].ack;
dsr_info->rsp->hdr.err = err < 0 ? -err : 0;
- pr_dbg("rsp->hdr.err=%d\n", dsr_info->rsp->hdr.err);
+
+ trace_pvrdma_exec_cmd(dsr_info->req->hdr.cmd, dsr_info->rsp->hdr.err);
out:
set_reg_val(dev, PVRDMA_REG_ERR, err);
diff --git a/hw/rdma/vmw/pvrdma_dev_ring.c b/hw/rdma/vmw/pvrdma_dev_ring.c
index e8e5b502f6..d7bc7f5ccc 100644
--- a/hw/rdma/vmw/pvrdma_dev_ring.c
+++ b/hw/rdma/vmw/pvrdma_dev_ring.c
@@ -17,6 +17,8 @@
#include "hw/pci/pci.h"
#include "cpu.h"
+#include "trace.h"
+
#include "../rdma_utils.h"
#include "standard-headers/drivers/infiniband/hw/vmw_pvrdma/pvrdma_ring.h"
#include "pvrdma_dev_ring.h"
@@ -30,13 +32,10 @@ int pvrdma_ring_init(PvrdmaRing *ring, const char *name, PCIDevice *dev,
strncpy(ring->name, name, MAX_RING_NAME_SZ);
ring->name[MAX_RING_NAME_SZ - 1] = 0;
- pr_dbg("Initializing %s ring\n", ring->name);
ring->dev = dev;
ring->ring_state = ring_state;
ring->max_elems = max_elems;
ring->elem_sz = elem_sz;
- pr_dbg("ring->elem_sz=%zu\n", ring->elem_sz);
- pr_dbg("npages=%d\n", npages);
/* TODO: Give a moment to think if we want to redo driver settings
atomic_set(&ring->ring_state->prod_tail, 0);
atomic_set(&ring->ring_state->cons_head, 0);
@@ -46,14 +45,14 @@ int pvrdma_ring_init(PvrdmaRing *ring, const char *name, PCIDevice *dev,
for (i = 0; i < npages; i++) {
if (!tbl[i]) {
- pr_err("npages=%ld but tbl[%d] is NULL\n", (long)npages, i);
+ rdma_error_report("npages=%d but tbl[%d] is NULL", npages, i);
continue;
}
ring->pages[i] = rdma_pci_dma_map(dev, tbl[i], TARGET_PAGE_SIZE);
if (!ring->pages[i]) {
rc = -ENOMEM;
- pr_dbg("Failed to map to page %d\n", i);
+ rdma_error_report("Failed to map to page %d in ring %s", i, name);
goto out_free;
}
memset(ring->pages[i], 0, TARGET_PAGE_SIZE);
@@ -78,7 +77,7 @@ void *pvrdma_ring_next_elem_read(PvrdmaRing *ring)
e = pvrdma_idx_ring_has_data(ring->ring_state, ring->max_elems, &idx);
if (e <= 0) {
- pr_dbg("No more data in ring\n");
+ trace_pvrdma_ring_next_elem_read_no_data(ring->name);
return NULL;
}
@@ -89,11 +88,6 @@ void *pvrdma_ring_next_elem_read(PvrdmaRing *ring)
void pvrdma_ring_read_inc(PvrdmaRing *ring)
{
pvrdma_idx_ring_inc(&ring->ring_state->cons_head, ring->max_elems);
- /*
- pr_dbg("%s: t=%d, h=%d, m=%ld\n", ring->name,
- ring->ring_state->prod_tail, ring->ring_state->cons_head,
- ring->max_elems);
- */
}
void *pvrdma_ring_next_elem_write(PvrdmaRing *ring)
@@ -103,13 +97,13 @@ void *pvrdma_ring_next_elem_write(PvrdmaRing *ring)
idx = pvrdma_idx_ring_has_space(ring->ring_state, ring->max_elems, &tail);
if (idx <= 0) {
- pr_dbg("CQ is full\n");
+ rdma_error_report("CQ is full");
return NULL;
}
idx = pvrdma_idx(&ring->ring_state->prod_tail, ring->max_elems);
if (idx < 0 || tail != idx) {
- pr_dbg("invalid idx\n");
+ rdma_error_report("Invalid idx %d", idx);
return NULL;
}
@@ -120,11 +114,6 @@ void *pvrdma_ring_next_elem_write(PvrdmaRing *ring)
void pvrdma_ring_write_inc(PvrdmaRing *ring)
{
pvrdma_idx_ring_inc(&ring->ring_state->prod_tail, ring->max_elems);
- /*
- pr_dbg("%s: t=%d, h=%d, m=%ld\n", ring->name,
- ring->ring_state->prod_tail, ring->ring_state->cons_head,
- ring->max_elems);
- */
}
void pvrdma_ring_free(PvrdmaRing *ring)
@@ -137,7 +126,6 @@ void pvrdma_ring_free(PvrdmaRing *ring)
return;
}
- pr_dbg("ring->npages=%d\n", ring->npages);
while (ring->npages--) {
rdma_pci_dma_unmap(ring->dev, ring->pages[ring->npages],
TARGET_PAGE_SIZE);
diff --git a/hw/rdma/vmw/pvrdma_main.c b/hw/rdma/vmw/pvrdma_main.c
index d2bdb5ba8c..b6061f4b6e 100644
--- a/hw/rdma/vmw/pvrdma_main.c
+++ b/hw/rdma/vmw/pvrdma_main.c
@@ -69,25 +69,22 @@ static int init_dev_ring(PvrdmaRing *ring, struct pvrdma_ring **ring_state,
uint64_t *dir, *tbl;
int rc = 0;
- pr_dbg("Initializing device ring %s\n", name);
- pr_dbg("pdir_dma=0x%llx\n", (long long unsigned int)dir_addr);
- pr_dbg("num_pages=%d\n", num_pages);
dir = rdma_pci_dma_map(pci_dev, dir_addr, TARGET_PAGE_SIZE);
if (!dir) {
- pr_err("Failed to map to page directory\n");
+ rdma_error_report("Failed to map to page directory (ring %s)", name);
rc = -ENOMEM;
goto out;
}
tbl = rdma_pci_dma_map(pci_dev, dir[0], TARGET_PAGE_SIZE);
if (!tbl) {
- pr_err("Failed to map to page table\n");
+ rdma_error_report("Failed to map to page table (ring %s)", name);
rc = -ENOMEM;
goto out_free_dir;
}
*ring_state = rdma_pci_dma_map(pci_dev, tbl[0], TARGET_PAGE_SIZE);
if (!*ring_state) {
- pr_err("Failed to map to ring state\n");
+ rdma_error_report("Failed to map to ring state (ring %s)", name);
rc = -ENOMEM;
goto out_free_tbl;
}
@@ -100,7 +97,6 @@ static int init_dev_ring(PvrdmaRing *ring, struct pvrdma_ring **ring_state,
sizeof(struct pvrdma_cqne),
(dma_addr_t *)&tbl[1], (dma_addr_t)num_pages - 1);
if (rc) {
- pr_err("Failed to initialize ring\n");
rc = -ENOMEM;
goto out_free_ring_state;
}
@@ -155,11 +151,10 @@ static int load_dsr(PVRDMADev *dev)
free_dsr(dev);
/* Map to DSR */
- pr_dbg("dsr_dma=0x%llx\n", (long long unsigned int)dev->dsr_info.dma);
dev->dsr_info.dsr = rdma_pci_dma_map(pci_dev, dev->dsr_info.dma,
sizeof(struct pvrdma_device_shared_region));
if (!dev->dsr_info.dsr) {
- pr_err("Failed to map to DSR\n");
+ rdma_error_report("Failed to map to DSR");
rc = -ENOMEM;
goto out;
}
@@ -169,21 +164,19 @@ static int load_dsr(PVRDMADev *dev)
dsr = dsr_info->dsr;
/* Map to command slot */
- pr_dbg("cmd_dma=0x%llx\n", (long long unsigned int)dsr->cmd_slot_dma);
dsr_info->req = rdma_pci_dma_map(pci_dev, dsr->cmd_slot_dma,
sizeof(union pvrdma_cmd_req));
if (!dsr_info->req) {
- pr_err("Failed to map to command slot address\n");
+ rdma_error_report("Failed to map to command slot address");
rc = -ENOMEM;
goto out_free_dsr;
}
/* Map to response slot */
- pr_dbg("rsp_dma=0x%llx\n", (long long unsigned int)dsr->resp_slot_dma);
dsr_info->rsp = rdma_pci_dma_map(pci_dev, dsr->resp_slot_dma,
sizeof(union pvrdma_cmd_resp));
if (!dsr_info->rsp) {
- pr_err("Failed to map to response slot address\n");
+ rdma_error_report("Failed to map to response slot address");
rc = -ENOMEM;
goto out_free_req;
}
@@ -193,7 +186,6 @@ static int load_dsr(PVRDMADev *dev)
pci_dev, dsr->cq_ring_pages.pdir_dma,
dsr->cq_ring_pages.num_pages);
if (rc) {
- pr_err("Failed to map to initialize CQ ring\n");
rc = -ENOMEM;
goto out_free_rsp;
}
@@ -203,7 +195,6 @@ static int load_dsr(PVRDMADev *dev)
"dev_async", pci_dev, dsr->async_ring_pages.pdir_dma,
dsr->async_ring_pages.num_pages);
if (rc) {
- pr_err("Failed to map to initialize event ring\n");
rc = -ENOMEM;
goto out_free_rsp;
}
@@ -230,24 +221,15 @@ static void init_dsr_dev_caps(PVRDMADev *dev)
struct pvrdma_device_shared_region *dsr;
if (dev->dsr_info.dsr == NULL) {
- pr_err("Can't initialized DSR\n");
+ rdma_error_report("Can't initialized DSR");
return;
}
dsr = dev->dsr_info.dsr;
-
dsr->caps.fw_ver = PVRDMA_FW_VERSION;
- pr_dbg("fw_ver=0x%" PRIx64 "\n", dsr->caps.fw_ver);
-
dsr->caps.mode = PVRDMA_DEVICE_MODE_ROCE;
- pr_dbg("mode=%d\n", dsr->caps.mode);
-
dsr->caps.gid_types |= PVRDMA_GID_TYPE_FLAG_ROCE_V1;
- pr_dbg("gid_types=0x%x\n", dsr->caps.gid_types);
-
dsr->caps.max_uar = RDMA_BAR2_UAR_SIZE;
- pr_dbg("max_uar=%d\n", dsr->caps.max_uar);
-
dsr->caps.max_mr_size = dev->dev_attr.max_mr_size;
dsr->caps.max_qp = dev->dev_attr.max_qp;
dsr->caps.max_qp_wr = dev->dev_attr.max_qp_wr;
@@ -257,23 +239,11 @@ static void init_dsr_dev_caps(PVRDMADev *dev)
dsr->caps.max_mr = dev->dev_attr.max_mr;
dsr->caps.max_pd = dev->dev_attr.max_pd;
dsr->caps.max_ah = dev->dev_attr.max_ah;
-
dsr->caps.gid_tbl_len = MAX_GIDS;
- pr_dbg("gid_tbl_len=%d\n", dsr->caps.gid_tbl_len);
-
dsr->caps.sys_image_guid = 0;
- pr_dbg("sys_image_guid=%" PRIx64 "\n", dsr->caps.sys_image_guid);
-
dsr->caps.node_guid = dev->node_guid;
- pr_dbg("node_guid=%" PRIx64 "\n", be64_to_cpu(dsr->caps.node_guid));
-
dsr->caps.phys_port_cnt = MAX_PORTS;
- pr_dbg("phys_port_cnt=%d\n", dsr->caps.phys_port_cnt);
-
dsr->caps.max_pkeys = MAX_PKEYS;
- pr_dbg("max_pkeys=%d\n", dsr->caps.max_pkeys);
-
- pr_dbg("Initialized\n");
}
static void uninit_msix(PCIDevice *pdev, int used_vectors)
@@ -319,9 +289,6 @@ static void pvrdma_fini(PCIDevice *pdev)
{
PVRDMADev *dev = PVRDMA_DEV(pdev);
- pr_dbg("Closing device %s %x.%x\n", pdev->name, PCI_SLOT(pdev->devfn),
- PCI_FUNC(pdev->devfn));
-
pvrdma_qp_ops_fini();
rdma_rm_fini(&dev->rdma_dev_res, &dev->backend_dev,
@@ -335,8 +302,8 @@ static void pvrdma_fini(PCIDevice *pdev)
uninit_msix(pdev, RDMA_MAX_INTRS);
}
- pr_dbg("Device %s %x.%x is down\n", pdev->name, PCI_SLOT(pdev->devfn),
- PCI_FUNC(pdev->devfn));
+ rdma_info_report("Device %s %x.%x is down", pdev->name,
+ PCI_SLOT(pdev->devfn), PCI_FUNC(pdev->devfn));
}
static void pvrdma_stop(PVRDMADev *dev)
@@ -353,32 +320,26 @@ static void activate_device(PVRDMADev *dev)
{
pvrdma_start(dev);
set_reg_val(dev, PVRDMA_REG_ERR, 0);
- pr_dbg("Device activated\n");
}
static int unquiesce_device(PVRDMADev *dev)
{
- pr_dbg("Device unquiesced\n");
return 0;
}
static void reset_device(PVRDMADev *dev)
{
pvrdma_stop(dev);
-
- pr_dbg("Device reset complete\n");
}
-static uint64_t regs_read(void *opaque, hwaddr addr, unsigned size)
+static uint64_t pvrdma_regs_read(void *opaque, hwaddr addr, unsigned size)
{
PVRDMADev *dev = opaque;
uint32_t val;
- /* pr_dbg("addr=0x%lx, size=%d\n", addr, size); */
-
if (get_reg_val(dev, addr, &val)) {
- pr_dbg("Error trying to read REG value from address 0x%x\n",
- (uint32_t)addr);
+ rdma_error_report("Failed to read REG value from address 0x%x",
+ (uint32_t)addr);
return -EINVAL;
}
@@ -387,25 +348,24 @@ static uint64_t regs_read(void *opaque, hwaddr addr, unsigned size)
return val;
}
-static void regs_write(void *opaque, hwaddr addr, uint64_t val, unsigned size)
+static void pvrdma_regs_write(void *opaque, hwaddr addr, uint64_t val,
+ unsigned size)
{
PVRDMADev *dev = opaque;
- /* pr_dbg("addr=0x%lx, val=0x%x, size=%d\n", addr, (uint32_t)val, size); */
-
if (set_reg_val(dev, addr, val)) {
- pr_err("Fail to set REG value, addr=0x%" PRIx64 ", val=0x%" PRIx64 "\n",
- addr, val);
+ rdma_error_report("Failed to set REG value, addr=0x%"PRIx64 ", val=0x%"PRIx64,
+ addr, val);
return;
}
- trace_pvrdma_regs_write(addr, val);
-
switch (addr) {
case PVRDMA_REG_DSRLOW:
+ trace_pvrdma_regs_write(addr, val, "DSRLOW", "");
dev->dsr_info.dma = val;
break;
case PVRDMA_REG_DSRHIGH:
+ trace_pvrdma_regs_write(addr, val, "DSRHIGH", "");
dev->dsr_info.dma |= val << 32;
load_dsr(dev);
init_dsr_dev_caps(dev);
@@ -413,23 +373,27 @@ static void regs_write(void *opaque, hwaddr addr, uint64_t val, unsigned size)
case PVRDMA_REG_CTL:
switch (val) {
case PVRDMA_DEVICE_CTL_ACTIVATE:
+ trace_pvrdma_regs_write(addr, val, "CTL", "ACTIVATE");
activate_device(dev);
break;
case PVRDMA_DEVICE_CTL_UNQUIESCE:
+ trace_pvrdma_regs_write(addr, val, "CTL", "UNQUIESCE");
unquiesce_device(dev);
break;
case PVRDMA_DEVICE_CTL_RESET:
+ trace_pvrdma_regs_write(addr, val, "CTL", "URESET");
reset_device(dev);
break;
}
break;
case PVRDMA_REG_IMR:
- pr_dbg("Interrupt mask=0x%" PRIx64 "\n", val);
+ trace_pvrdma_regs_write(addr, val, "INTR_MASK", "");
dev->interrupt_mask = val;
break;
case PVRDMA_REG_REQUEST:
if (val == 0) {
- execute_command(dev);
+ trace_pvrdma_regs_write(addr, val, "REQUEST", "");
+ pvrdma_exec_cmd(dev);
}
break;
default:
@@ -438,8 +402,8 @@ static void regs_write(void *opaque, hwaddr addr, uint64_t val, unsigned size)
}
static const MemoryRegionOps regs_ops = {
- .read = regs_read,
- .write = regs_write,
+ .read = pvrdma_regs_read,
+ .write = pvrdma_regs_write,
.endianness = DEVICE_LITTLE_ENDIAN,
.impl = {
.min_access_size = sizeof(uint32_t),
@@ -447,54 +411,58 @@ static const MemoryRegionOps regs_ops = {
},
};
-static uint64_t uar_read(void *opaque, hwaddr addr, unsigned size)
+static uint64_t pvrdma_uar_read(void *opaque, hwaddr addr, unsigned size)
{
return 0xffffffff;
}
-static void uar_write(void *opaque, hwaddr addr, uint64_t val, unsigned size)
+static void pvrdma_uar_write(void *opaque, hwaddr addr, uint64_t val,
+ unsigned size)
{
PVRDMADev *dev = opaque;
- /* pr_dbg("addr=0x%lx, val=0x%x, size=%d\n", addr, (uint32_t)val, size); */
-
switch (addr & 0xFFF) { /* Mask with 0xFFF as each UC gets page */
case PVRDMA_UAR_QP_OFFSET:
- pr_dbg("UAR QP command, addr=0x%" PRIx64 ", val=0x%" PRIx64 "\n",
- (uint64_t)addr, val);
if (val & PVRDMA_UAR_QP_SEND) {
+ trace_pvrdma_uar_write(addr, val, "QP", "SEND",
+ val & PVRDMA_UAR_HANDLE_MASK, 0);
pvrdma_qp_send(dev, val & PVRDMA_UAR_HANDLE_MASK);
}
if (val & PVRDMA_UAR_QP_RECV) {
+ trace_pvrdma_uar_write(addr, val, "QP", "RECV",
+ val & PVRDMA_UAR_HANDLE_MASK, 0);
pvrdma_qp_recv(dev, val & PVRDMA_UAR_HANDLE_MASK);
}
break;
case PVRDMA_UAR_CQ_OFFSET:
- /* pr_dbg("UAR CQ cmd, addr=0x%x, val=0x%lx\n", (uint32_t)addr, val); */
if (val & PVRDMA_UAR_CQ_ARM) {
+ trace_pvrdma_uar_write(addr, val, "CQ", "ARM",
+ val & PVRDMA_UAR_HANDLE_MASK,
+ !!(val & PVRDMA_UAR_CQ_ARM_SOL));
rdma_rm_req_notify_cq(&dev->rdma_dev_res,
val & PVRDMA_UAR_HANDLE_MASK,
!!(val & PVRDMA_UAR_CQ_ARM_SOL));
}
if (val & PVRDMA_UAR_CQ_ARM_SOL) {
- pr_dbg("UAR_CQ_ARM_SOL (%" PRIx64 ")\n",
- val & PVRDMA_UAR_HANDLE_MASK);
+ trace_pvrdma_uar_write(addr, val, "CQ", "ARMSOL - not supported", 0,
+ 0);
}
if (val & PVRDMA_UAR_CQ_POLL) {
- pr_dbg("UAR_CQ_POLL (%" PRIx64 ")\n", val & PVRDMA_UAR_HANDLE_MASK);
+ trace_pvrdma_uar_write(addr, val, "CQ", "POLL",
+ val & PVRDMA_UAR_HANDLE_MASK, 0);
pvrdma_cq_poll(&dev->rdma_dev_res, val & PVRDMA_UAR_HANDLE_MASK);
}
break;
default:
- pr_err("Unsupported command, addr=0x%" PRIx64 ", val=0x%" PRIx64 "\n",
- addr, val);
+ rdma_error_report("Unsupported command, addr=0x%"PRIx64", val=0x%"PRIx64,
+ addr, val);
break;
}
}
static const MemoryRegionOps uar_ops = {
- .read = uar_read,
- .write = uar_write,
+ .read = pvrdma_uar_read,
+ .write = pvrdma_uar_write,
.endianness = DEVICE_LITTLE_ENDIAN,
.impl = {
.min_access_size = sizeof(uint32_t),
@@ -551,11 +519,9 @@ static void init_dev_caps(PVRDMADev *dev)
(wr_sz + sizeof(struct pvrdma_sge) *
dev->dev_attr.max_sge) - TARGET_PAGE_SIZE;
/* First page is ring state ^^^^ */
- pr_dbg("max_qp_wr=%d\n", dev->dev_attr.max_qp_wr);
dev->dev_attr.max_cqe = pg_tbl_bytes / sizeof(struct pvrdma_cqe) -
TARGET_PAGE_SIZE; /* First page is ring state */
- pr_dbg("max_cqe=%d\n", dev->dev_attr.max_cqe);
}
static int pvrdma_check_ram_shared(Object *obj, void *opaque)
@@ -585,10 +551,8 @@ static void pvrdma_realize(PCIDevice *pdev, Error **errp)
bool ram_shared = false;
PCIDevice *func0;
- init_pr_dbg();
-
- pr_dbg("Initializing device %s %x.%x\n", pdev->name,
- PCI_SLOT(pdev->devfn), PCI_FUNC(pdev->devfn));
+ rdma_info_report("Initializing device %s %x.%x", pdev->name,
+ PCI_SLOT(pdev->devfn), PCI_FUNC(pdev->devfn));
if (TARGET_PAGE_SIZE != getpagesize()) {
error_setg(errp, "Target page size must be the same as host page size");
@@ -598,8 +562,6 @@ static void pvrdma_realize(PCIDevice *pdev, Error **errp)
func0 = pci_get_function_0(pdev);
/* Break if not vmxnet3 device in slot 0 */
if (strcmp(object_get_typename(&func0->qdev.parent_obj), TYPE_VMXNET3)) {
- pr_dbg("func0 type is %s\n",
- object_get_typename(&func0->qdev.parent_obj));
error_setg(errp, "Device on %x.0 must be %s", PCI_SLOT(pdev->devfn),
TYPE_VMXNET3);
return;
@@ -656,7 +618,7 @@ static void pvrdma_realize(PCIDevice *pdev, Error **errp)
out:
if (rc) {
pvrdma_fini(pdev);
- error_append_hint(errp, "Device fail to load\n");
+ error_append_hint(errp, "Device failed to load\n");
}
}
diff --git a/hw/rdma/vmw/pvrdma_qp_ops.c b/hw/rdma/vmw/pvrdma_qp_ops.c
index ce5a60e184..16db726dac 100644
--- a/hw/rdma/vmw/pvrdma_qp_ops.c
+++ b/hw/rdma/vmw/pvrdma_qp_ops.c
@@ -19,6 +19,8 @@
#include "../rdma_rm.h"
#include "../rdma_backend.h"
+#include "trace.h"
+
#include "pvrdma.h"
#include "standard-headers/rdma/vmw_pvrdma-abi.h"
#include "pvrdma_qp_ops.h"
@@ -55,18 +57,14 @@ static int pvrdma_post_cqe(PVRDMADev *dev, uint32_t cq_handle,
RdmaRmCQ *cq = rdma_rm_get_cq(&dev->rdma_dev_res, cq_handle);
if (unlikely(!cq)) {
- pr_dbg("Invalid cqn %d\n", cq_handle);
return -EINVAL;
}
ring = (PvrdmaRing *)cq->opaque;
- pr_dbg("ring=%p\n", ring);
/* Step #1: Put CQE on CQ ring */
- pr_dbg("Writing CQE\n");
cqe1 = pvrdma_ring_next_elem_write(ring);
if (unlikely(!cqe1)) {
- pr_dbg("No CQEs in ring\n");
return -EINVAL;
}
@@ -80,19 +78,13 @@ static int pvrdma_post_cqe(PVRDMADev *dev, uint32_t cq_handle,
cqe1->wc_flags = wc->wc_flags;
cqe1->vendor_err = wc->vendor_err;
- pr_dbg("wr_id=%" PRIx64 "\n", cqe1->wr_id);
- pr_dbg("qp=0x%lx\n", cqe1->qp);
- pr_dbg("opcode=%d\n", cqe1->opcode);
- pr_dbg("status=%d\n", cqe1->status);
- pr_dbg("byte_len=%d\n", cqe1->byte_len);
- pr_dbg("src_qp=%d\n", cqe1->src_qp);
- pr_dbg("wc_flags=%d\n", cqe1->wc_flags);
- pr_dbg("vendor_err=%d\n", cqe1->vendor_err);
+ trace_pvrdma_post_cqe(cq_handle, cq->notify, cqe1->wr_id, cqe1->qp,
+ cqe1->opcode, cqe1->status, cqe1->byte_len,
+ cqe1->src_qp, cqe1->wc_flags, cqe1->vendor_err);
pvrdma_ring_write_inc(ring);
/* Step #2: Put CQ number on dsr completion ring */
- pr_dbg("Writing CQNE\n");
cqne = pvrdma_ring_next_elem_write(&dev->dsr_info.cq);
if (unlikely(!cqne)) {
return -EINVAL;
@@ -101,7 +93,6 @@ static int pvrdma_post_cqe(PVRDMADev *dev, uint32_t cq_handle,
cqne->info = cq_handle;
pvrdma_ring_write_inc(&dev->dsr_info.cq);
- pr_dbg("cq->notify=%d\n", cq->notify);
if (cq->notify != CNT_CLEAR) {
if (cq->notify == CNT_ARM) {
cq->notify = CNT_CLEAR;
@@ -151,23 +142,17 @@ void pvrdma_qp_send(PVRDMADev *dev, uint32_t qp_handle)
int sgid_idx;
union ibv_gid *sgid;
- pr_dbg("qp_handle=0x%x\n", qp_handle);
-
qp = rdma_rm_get_qp(&dev->rdma_dev_res, qp_handle);
if (unlikely(!qp)) {
- pr_dbg("Invalid qpn\n");
return;
}
ring = (PvrdmaRing *)qp->opaque;
- pr_dbg("sring=%p\n", ring);
wqe = (struct PvrdmaSqWqe *)pvrdma_ring_next_elem_read(ring);
while (wqe) {
CompHandlerCtx *comp_ctx;
- pr_dbg("wr_id=%" PRIx64 "\n", wqe->hdr.wr_id);
-
/* Prepare CQE */
comp_ctx = g_malloc(sizeof(CompHandlerCtx));
comp_ctx->dev = dev;
@@ -178,26 +163,25 @@ void pvrdma_qp_send(PVRDMADev *dev, uint32_t qp_handle)
sgid = rdma_rm_get_gid(&dev->rdma_dev_res, wqe->hdr.wr.ud.av.gid_index);
if (!sgid) {
- pr_dbg("Fail to get gid for idx %d\n", wqe->hdr.wr.ud.av.gid_index);
+ rdma_error_report("Failed to get gid for idx %d",
+ wqe->hdr.wr.ud.av.gid_index);
complete_with_error(VENDOR_ERR_INV_GID_IDX, comp_ctx);
continue;
}
- pr_dbg("sgid_id=%d, sgid=0x%llx\n", wqe->hdr.wr.ud.av.gid_index,
- sgid->global.interface_id);
sgid_idx = rdma_rm_get_backend_gid_index(&dev->rdma_dev_res,
&dev->backend_dev,
wqe->hdr.wr.ud.av.gid_index);
if (sgid_idx <= 0) {
- pr_dbg("Fail to get bk sgid_idx for sgid_idx %d\n",
- wqe->hdr.wr.ud.av.gid_index);
+ rdma_error_report("Failed to get bk sgid_idx for sgid_idx %d",
+ wqe->hdr.wr.ud.av.gid_index);
complete_with_error(VENDOR_ERR_INV_GID_IDX, comp_ctx);
continue;
}
if (wqe->hdr.num_sge > dev->dev_attr.max_sge) {
- pr_dbg("Invalid num_sge=%d (max %d)\n", wqe->hdr.num_sge,
- dev->dev_attr.max_sge);
+ rdma_error_report("Invalid num_sge=%d (max %d)", wqe->hdr.num_sge,
+ dev->dev_attr.max_sge);
complete_with_error(VENDOR_ERR_INV_NUM_SGE, comp_ctx);
continue;
}
@@ -221,23 +205,17 @@ void pvrdma_qp_recv(PVRDMADev *dev, uint32_t qp_handle)
PvrdmaRqWqe *wqe;
PvrdmaRing *ring;
- pr_dbg("qp_handle=0x%x\n", qp_handle);
-
qp = rdma_rm_get_qp(&dev->rdma_dev_res, qp_handle);
if (unlikely(!qp)) {
- pr_dbg("Invalid qpn\n");
return;
}
ring = &((PvrdmaRing *)qp->opaque)[1];
- pr_dbg("rring=%p\n", ring);
wqe = (struct PvrdmaRqWqe *)pvrdma_ring_next_elem_read(ring);
while (wqe) {
CompHandlerCtx *comp_ctx;
- pr_dbg("wr_id=%" PRIx64 "\n", wqe->hdr.wr_id);
-
/* Prepare CQE */
comp_ctx = g_malloc(sizeof(CompHandlerCtx));
comp_ctx->dev = dev;
@@ -247,8 +225,8 @@ void pvrdma_qp_recv(PVRDMADev *dev, uint32_t qp_handle)
comp_ctx->cqe.opcode = IBV_WC_RECV;
if (wqe->hdr.num_sge > dev->dev_attr.max_sge) {
- pr_dbg("Invalid num_sge=%d (max %d)\n", wqe->hdr.num_sge,
- dev->dev_attr.max_sge);
+ rdma_error_report("Invalid num_sge=%d (max %d)", wqe->hdr.num_sge,
+ dev->dev_attr.max_sge);
complete_with_error(VENDOR_ERR_INV_NUM_SGE, comp_ctx);
continue;
}
@@ -270,7 +248,6 @@ void pvrdma_cq_poll(RdmaDeviceResources *dev_res, uint32_t cq_handle)
cq = rdma_rm_get_cq(dev_res, cq_handle);
if (!cq) {
- pr_dbg("Invalid CQ# %d\n", cq_handle);
return;
}
diff --git a/hw/rdma/vmw/trace-events b/hw/rdma/vmw/trace-events
index b3f9e2b19f..0122266ad7 100644
--- a/hw/rdma/vmw/trace-events
+++ b/hw/rdma/vmw/trace-events
@@ -1,5 +1,17 @@
# See docs/tracing.txt for syntax documentation.
# hw/rdma/vmw/pvrdma_main.c
-pvrdma_regs_read(uint64_t addr, uint64_t val) "regs[0x%"PRIx64"] = 0x%"PRIx64
-pvrdma_regs_write(uint64_t addr, uint64_t val) "regs[0x%"PRIx64"] = 0x%"PRIx64
+pvrdma_regs_read(uint64_t addr, uint64_t val) "pvrdma.regs[0x%"PRIx64"]=0x%"PRIx64
+pvrdma_regs_write(uint64_t addr, uint64_t val, const char *reg_name, const char *val_name) "pvrdma.regs[0x%"PRIx64"]=0x%"PRIx64" (%s %s)"
+pvrdma_uar_write(uint64_t addr, uint64_t val, const char *reg_name, const char *val_name, int val1, int val2) "uar[0x%"PRIx64"]=0x%"PRIx64" (cls=%s, op=%s, obj=%d, val=%d)"
+
+# hw/rdma/vmw/pvrdma_cmd.c
+pvrdma_map_to_pdir_host_virt(void *vfirst, void *vremaped) "mremap %p -> %p"
+pvrdma_map_to_pdir_next_page(int page_idx, void *vnext, void *vremaped) "mremap [%d] %p -> %p"
+pvrdma_exec_cmd(int cmd, int err) "cmd=%d, err=%d"
+
+# hw/rdma/vmw/pvrdma_dev_ring.c
+pvrdma_ring_next_elem_read_no_data(char *ring_name) "pvrdma_ring %s is empty"
+
+# hw/rdma/vmw/pvrdma_qp_ops.c
+pvrdma_post_cqe(uint32_t cq_handle, int notify, uint64_t wr_id, uint64_t qpn, uint32_t op_code, uint32_t status, uint32_t byte_len, uint32_t src_qp, uint32_t wc_flags, uint32_t vendor_err) "cq_handle=%d, notify=%d, wr_id=0x%"PRIx64", qpn=0x%"PRIx64", opcode=%d, status=%d, byte_len=%d, src_qp=%d, wc_flags=%d, vendor_err=%d"
--
2.17.2
^ permalink raw reply related [flat|nested] 17+ messages in thread
* [Qemu-devel] [PATCH v4 2/9] hw/rdma: Introduce protected qlist
2019-03-03 20:33 [Qemu-devel] [PATCH v4 0/9] Misc fixes to pvrdma device Yuval Shaia
2019-03-03 20:33 ` [Qemu-devel] [PATCH v4 1/9] hw/rdma: Switch to generic error reporting way Yuval Shaia
@ 2019-03-03 20:33 ` Yuval Shaia
2019-03-03 20:33 ` [Qemu-devel] [PATCH v4 3/9] hw/rdma: Protect against concurrent execution of poll_cq Yuval Shaia
` (6 subsequent siblings)
8 siblings, 0 replies; 17+ messages in thread
From: Yuval Shaia @ 2019-03-03 20:33 UTC (permalink / raw)
To: dgilbert, yuval.shaia, marcel.apfelbaum, armbru, qemu-devel
To make code more readable move handling of protected list to a
rdma_utils
Signed-off-by: Yuval Shaia <yuval.shaia@oracle.com>
Reviewed-by: Marcel Apfelbaum <marcel.apfelbaum@gmail.com>
---
hw/rdma/rdma_backend.c | 20 +++++--------------
hw/rdma/rdma_backend_defs.h | 8 ++------
hw/rdma/rdma_utils.c | 39 +++++++++++++++++++++++++++++++++++++
hw/rdma/rdma_utils.h | 9 +++++++++
4 files changed, 55 insertions(+), 21 deletions(-)
diff --git a/hw/rdma/rdma_backend.c b/hw/rdma/rdma_backend.c
index 24bac00a20..0ed14751be 100644
--- a/hw/rdma/rdma_backend.c
+++ b/hw/rdma/rdma_backend.c
@@ -527,9 +527,7 @@ static unsigned int save_mad_recv_buffer(RdmaBackendDev *backend_dev,
bctx->up_ctx = ctx;
bctx->sge = *sge;
- qemu_mutex_lock(&backend_dev->recv_mads_list.lock);
- qlist_append_int(backend_dev->recv_mads_list.list, bctx_id);
- qemu_mutex_unlock(&backend_dev->recv_mads_list.lock);
+ rdma_protected_qlist_append_int64(&backend_dev->recv_mads_list, bctx_id);
return 0;
}
@@ -913,23 +911,19 @@ static inline void build_mad_hdr(struct ibv_grh *grh, union ibv_gid *sgid,
static void process_incoming_mad_req(RdmaBackendDev *backend_dev,
RdmaCmMuxMsg *msg)
{
- QObject *o_ctx_id;
unsigned long cqe_ctx_id;
BackendCtx *bctx;
char *mad;
trace_mad_message("recv", msg->umad.mad, msg->umad_len);
- qemu_mutex_lock(&backend_dev->recv_mads_list.lock);
- o_ctx_id = qlist_pop(backend_dev->recv_mads_list.list);
- qemu_mutex_unlock(&backend_dev->recv_mads_list.lock);
- if (!o_ctx_id) {
+ cqe_ctx_id = rdma_protected_qlist_pop_int64(&backend_dev->recv_mads_list);
+ if (cqe_ctx_id == -ENOENT) {
rdma_warn_report("No more free MADs buffers, waiting for a while");
sleep(THR_POLL_TO);
return;
}
- cqe_ctx_id = qnum_get_uint(qobject_to(QNum, o_ctx_id));
bctx = rdma_rm_get_cqe_ctx(backend_dev->rdma_dev_res, cqe_ctx_id);
if (unlikely(!bctx)) {
rdma_error_report("No matching ctx for req %ld", cqe_ctx_id);
@@ -994,8 +988,7 @@ static int mad_init(RdmaBackendDev *backend_dev, CharBackend *mad_chr_be)
return -EIO;
}
- qemu_mutex_init(&backend_dev->recv_mads_list.lock);
- backend_dev->recv_mads_list.list = qlist_new();
+ rdma_protected_qlist_init(&backend_dev->recv_mads_list);
enable_rdmacm_mux_async(backend_dev);
@@ -1010,10 +1003,7 @@ static void mad_fini(RdmaBackendDev *backend_dev)
{
disable_rdmacm_mux_async(backend_dev);
qemu_chr_fe_disconnect(backend_dev->rdmacm_mux.chr_be);
- if (backend_dev->recv_mads_list.list) {
- qlist_destroy_obj(QOBJECT(backend_dev->recv_mads_list.list));
- qemu_mutex_destroy(&backend_dev->recv_mads_list.lock);
- }
+ rdma_protected_qlist_destroy(&backend_dev->recv_mads_list);
}
int rdma_backend_get_gid_index(RdmaBackendDev *backend_dev,
diff --git a/hw/rdma/rdma_backend_defs.h b/hw/rdma/rdma_backend_defs.h
index 15ae8b970e..a8c15b09ab 100644
--- a/hw/rdma/rdma_backend_defs.h
+++ b/hw/rdma/rdma_backend_defs.h
@@ -20,6 +20,7 @@
#include "chardev/char-fe.h"
#include <infiniband/verbs.h>
#include "contrib/rdmacm-mux/rdmacm-mux.h"
+#include "rdma_utils.h"
typedef struct RdmaDeviceResources RdmaDeviceResources;
@@ -30,11 +31,6 @@ typedef struct RdmaBackendThread {
bool is_running; /* Set by the thread to report its status */
} RdmaBackendThread;
-typedef struct RecvMadList {
- QemuMutex lock;
- QList *list;
-} RecvMadList;
-
typedef struct RdmaCmMux {
CharBackend *chr_be;
int can_receive;
@@ -48,7 +44,7 @@ typedef struct RdmaBackendDev {
struct ibv_context *context;
struct ibv_comp_channel *channel;
uint8_t port_num;
- RecvMadList recv_mads_list;
+ RdmaProtectedQList recv_mads_list;
RdmaCmMux rdmacm_mux;
} RdmaBackendDev;
diff --git a/hw/rdma/rdma_utils.c b/hw/rdma/rdma_utils.c
index b9f07fcda7..0a8abe572d 100644
--- a/hw/rdma/rdma_utils.c
+++ b/hw/rdma/rdma_utils.c
@@ -14,6 +14,8 @@
*/
#include "qemu/osdep.h"
+#include "qapi/qmp/qlist.h"
+#include "qapi/qmp/qnum.h"
#include "trace.h"
#include "rdma_utils.h"
@@ -51,3 +53,40 @@ void rdma_pci_dma_unmap(PCIDevice *dev, void *buffer, dma_addr_t len)
pci_dma_unmap(dev, buffer, len, DMA_DIRECTION_TO_DEVICE, 0);
}
}
+
+void rdma_protected_qlist_init(RdmaProtectedQList *list)
+{
+ qemu_mutex_init(&list->lock);
+ list->list = qlist_new();
+}
+
+void rdma_protected_qlist_destroy(RdmaProtectedQList *list)
+{
+ if (list->list) {
+ qlist_destroy_obj(QOBJECT(list->list));
+ qemu_mutex_destroy(&list->lock);
+ list->list = NULL;
+ }
+}
+
+void rdma_protected_qlist_append_int64(RdmaProtectedQList *list, int64_t value)
+{
+ qemu_mutex_lock(&list->lock);
+ qlist_append_int(list->list, value);
+ qemu_mutex_unlock(&list->lock);
+}
+
+int64_t rdma_protected_qlist_pop_int64(RdmaProtectedQList *list)
+{
+ QObject *obj;
+
+ qemu_mutex_lock(&list->lock);
+ obj = qlist_pop(list->list);
+ qemu_mutex_unlock(&list->lock);
+
+ if (!obj) {
+ return -ENOENT;
+ }
+
+ return qnum_get_uint(qobject_to(QNum, obj));
+}
diff --git a/hw/rdma/rdma_utils.h b/hw/rdma/rdma_utils.h
index acd148837f..a8bf1d4fec 100644
--- a/hw/rdma/rdma_utils.h
+++ b/hw/rdma/rdma_utils.h
@@ -29,8 +29,17 @@
#define rdma_info_report(fmt, ...) \
info_report("%s: " fmt, "rdma", ## __VA_ARGS__)
+typedef struct RdmaProtectedQList {
+ QemuMutex lock;
+ QList *list;
+} RdmaProtectedQList;
+
void *rdma_pci_dma_map(PCIDevice *dev, dma_addr_t addr, dma_addr_t plen);
void rdma_pci_dma_unmap(PCIDevice *dev, void *buffer, dma_addr_t len);
+void rdma_protected_qlist_init(RdmaProtectedQList *list);
+void rdma_protected_qlist_destroy(RdmaProtectedQList *list);
+void rdma_protected_qlist_append_int64(RdmaProtectedQList *list, int64_t value);
+int64_t rdma_protected_qlist_pop_int64(RdmaProtectedQList *list);
static inline void addrconf_addr_eui48(uint8_t *eui, const char *addr)
{
--
2.17.2
^ permalink raw reply related [flat|nested] 17+ messages in thread
* [Qemu-devel] [PATCH v4 3/9] hw/rdma: Protect against concurrent execution of poll_cq
2019-03-03 20:33 [Qemu-devel] [PATCH v4 0/9] Misc fixes to pvrdma device Yuval Shaia
2019-03-03 20:33 ` [Qemu-devel] [PATCH v4 1/9] hw/rdma: Switch to generic error reporting way Yuval Shaia
2019-03-03 20:33 ` [Qemu-devel] [PATCH v4 2/9] hw/rdma: Introduce protected qlist Yuval Shaia
@ 2019-03-03 20:33 ` Yuval Shaia
2019-03-03 20:33 ` [Qemu-devel] [PATCH v4 4/9] {hmp, hw/pvrdma}: Expose device internals via monitor interface Yuval Shaia
` (5 subsequent siblings)
8 siblings, 0 replies; 17+ messages in thread
From: Yuval Shaia @ 2019-03-03 20:33 UTC (permalink / raw)
To: dgilbert, yuval.shaia, marcel.apfelbaum, armbru, qemu-devel
The function rdma_poll_cq is called from two contexts - completion
handler thread which sense new completion on backend channel and
explicitly as result of guest issuing poll_cq command.
Add lock to protect against concurrent executions.
Signed-off-by: Yuval Shaia <yuval.shaia@oracle.com>
Reviewed-by: Marcel Apfelbaum <marcel.apfelbaum@gmail.com>
---
hw/rdma/rdma_backend.c | 2 ++
hw/rdma/rdma_rm.c | 4 ++++
hw/rdma/rdma_rm_defs.h | 1 +
3 files changed, 7 insertions(+)
diff --git a/hw/rdma/rdma_backend.c b/hw/rdma/rdma_backend.c
index 0ed14751be..9679b842d1 100644
--- a/hw/rdma/rdma_backend.c
+++ b/hw/rdma/rdma_backend.c
@@ -70,6 +70,7 @@ static void rdma_poll_cq(RdmaDeviceResources *rdma_dev_res, struct ibv_cq *ibcq)
BackendCtx *bctx;
struct ibv_wc wc[2];
+ qemu_mutex_lock(&rdma_dev_res->lock);
do {
ne = ibv_poll_cq(ibcq, ARRAY_SIZE(wc), wc);
@@ -89,6 +90,7 @@ static void rdma_poll_cq(RdmaDeviceResources *rdma_dev_res, struct ibv_cq *ibcq)
g_free(bctx);
}
} while (ne > 0);
+ qemu_mutex_unlock(&rdma_dev_res->lock);
if (ne < 0) {
rdma_error_report("ibv_poll_cq fail, rc=%d, errno=%d", ne, errno);
diff --git a/hw/rdma/rdma_rm.c b/hw/rdma/rdma_rm.c
index 5dab4a2189..14580ca379 100644
--- a/hw/rdma/rdma_rm.c
+++ b/hw/rdma/rdma_rm.c
@@ -618,12 +618,16 @@ int rdma_rm_init(RdmaDeviceResources *dev_res, struct ibv_device_attr *dev_attr,
init_ports(dev_res);
+ qemu_mutex_init(&dev_res->lock);
+
return 0;
}
void rdma_rm_fini(RdmaDeviceResources *dev_res, RdmaBackendDev *backend_dev,
const char *ifname)
{
+ qemu_mutex_destroy(&dev_res->lock);
+
fini_ports(dev_res, backend_dev, ifname);
res_tbl_free(&dev_res->uc_tbl);
diff --git a/hw/rdma/rdma_rm_defs.h b/hw/rdma/rdma_rm_defs.h
index 0ba61d1838..f0ee1f3072 100644
--- a/hw/rdma/rdma_rm_defs.h
+++ b/hw/rdma/rdma_rm_defs.h
@@ -105,6 +105,7 @@ typedef struct RdmaDeviceResources {
RdmaRmResTbl cq_tbl;
RdmaRmResTbl cqe_ctx_tbl;
GHashTable *qp_hash; /* Keeps mapping between real and emulated */
+ QemuMutex lock;
} RdmaDeviceResources;
#endif
--
2.17.2
^ permalink raw reply related [flat|nested] 17+ messages in thread
* [Qemu-devel] [PATCH v4 4/9] {hmp, hw/pvrdma}: Expose device internals via monitor interface
2019-03-03 20:33 [Qemu-devel] [PATCH v4 0/9] Misc fixes to pvrdma device Yuval Shaia
` (2 preceding siblings ...)
2019-03-03 20:33 ` [Qemu-devel] [PATCH v4 3/9] hw/rdma: Protect against concurrent execution of poll_cq Yuval Shaia
@ 2019-03-03 20:33 ` Yuval Shaia
2019-03-06 10:22 ` Yuval Shaia
` (2 more replies)
2019-03-03 20:33 ` [Qemu-devel] [PATCH v4 5/9] hw/rdma: Free all MAD receive buffers when device is closed Yuval Shaia
` (4 subsequent siblings)
8 siblings, 3 replies; 17+ messages in thread
From: Yuval Shaia @ 2019-03-03 20:33 UTC (permalink / raw)
To: dgilbert, yuval.shaia, marcel.apfelbaum, armbru, qemu-devel
Allow interrogating device internals through HMP interface.
The exposed indicators can be used for troubleshooting by developers or
sysadmin.
There is no need to expose these attributes to a management system (e.x.
libvirt) because (1) most of them are not "device-management' related
info and (2) there is no guarantee the interface is stable.
Signed-off-by: Yuval Shaia <yuval.shaia@oracle.com>
---
hmp-commands-info.hx | 14 ++++++++
hmp.c | 27 +++++++++++++++
hmp.h | 1 +
hw/rdma/Makefile.objs | 2 +-
hw/rdma/rdma_backend.c | 70 +++++++++++++++++++++++++++++---------
hw/rdma/rdma_hmp.c | 30 ++++++++++++++++
hw/rdma/rdma_rm.c | 60 ++++++++++++++++++++++++++++++++
hw/rdma/rdma_rm.h | 1 +
hw/rdma/rdma_rm_defs.h | 27 ++++++++++++++-
hw/rdma/vmw/pvrdma.h | 5 +++
hw/rdma/vmw/pvrdma_main.c | 21 ++++++++++++
include/hw/rdma/rdma_hmp.h | 40 ++++++++++++++++++++++
12 files changed, 279 insertions(+), 19 deletions(-)
create mode 100644 hw/rdma/rdma_hmp.c
create mode 100644 include/hw/rdma/rdma_hmp.h
diff --git a/hmp-commands-info.hx b/hmp-commands-info.hx
index cbee8b944d..c59444c461 100644
--- a/hmp-commands-info.hx
+++ b/hmp-commands-info.hx
@@ -202,6 +202,20 @@ STEXI
@item info pic
@findex info pic
Show PIC state.
+ETEXI
+
+ {
+ .name = "rdma",
+ .args_type = "",
+ .params = "",
+ .help = "show RDMA state",
+ .cmd = hmp_info_rdma,
+ },
+
+STEXI
+@item info rdma
+@findex info rdma
+Show RDMA state.
ETEXI
{
diff --git a/hmp.c b/hmp.c
index 1e006eeb49..68511b2441 100644
--- a/hmp.c
+++ b/hmp.c
@@ -51,6 +51,7 @@
#include "qemu/error-report.h"
#include "exec/ramlist.h"
#include "hw/intc/intc.h"
+#include "hw/rdma/rdma_hmp.h"
#include "migration/snapshot.h"
#include "migration/misc.h"
@@ -968,6 +969,32 @@ void hmp_info_pic(Monitor *mon, const QDict *qdict)
hmp_info_pic_foreach, mon);
}
+static int hmp_info_rdma_foreach(Object *obj, void *opaque)
+{
+ RdmaStatsProvider *rdma;
+ RdmaStatsProviderClass *k;
+ Monitor *mon = opaque;
+
+ if (object_dynamic_cast(obj, TYPE_RDMA_STATS_PROVIDER)) {
+ rdma = RDMA_STATS_PROVIDER(obj);
+ k = RDMA_STATS_PROVIDER_GET_CLASS(obj);
+ if (k->print_statistics) {
+ k->print_statistics(mon, rdma);
+ } else {
+ monitor_printf(mon, "RDMA statistics not available for %s.\n",
+ object_get_typename(obj));
+ }
+ }
+
+ return 0;
+}
+
+void hmp_info_rdma(Monitor *mon, const QDict *qdict)
+{
+ object_child_foreach_recursive(object_get_root(),
+ hmp_info_rdma_foreach, mon);
+}
+
void hmp_info_pci(Monitor *mon, const QDict *qdict)
{
PciInfoList *info_list, *info;
diff --git a/hmp.h b/hmp.h
index 5f1addcca2..666949afc3 100644
--- a/hmp.h
+++ b/hmp.h
@@ -36,6 +36,7 @@ void hmp_info_spice(Monitor *mon, const QDict *qdict);
void hmp_info_balloon(Monitor *mon, const QDict *qdict);
void hmp_info_irq(Monitor *mon, const QDict *qdict);
void hmp_info_pic(Monitor *mon, const QDict *qdict);
+void hmp_info_rdma(Monitor *mon, const QDict *qdict);
void hmp_info_pci(Monitor *mon, const QDict *qdict);
void hmp_info_block_jobs(Monitor *mon, const QDict *qdict);
void hmp_info_tpm(Monitor *mon, const QDict *qdict);
diff --git a/hw/rdma/Makefile.objs b/hw/rdma/Makefile.objs
index bd36cbf51c..dd59faff51 100644
--- a/hw/rdma/Makefile.objs
+++ b/hw/rdma/Makefile.objs
@@ -1,5 +1,5 @@
ifeq ($(CONFIG_PVRDMA),y)
-obj-$(CONFIG_PCI) += rdma_utils.o rdma_backend.o rdma_rm.o
+obj-$(CONFIG_PCI) += rdma_utils.o rdma_backend.o rdma_rm.o rdma_hmp.o
obj-$(CONFIG_PCI) += vmw/pvrdma_dev_ring.o vmw/pvrdma_cmd.o \
vmw/pvrdma_qp_ops.o vmw/pvrdma_main.o
endif
diff --git a/hw/rdma/rdma_backend.c b/hw/rdma/rdma_backend.c
index 9679b842d1..bc2fefcf93 100644
--- a/hw/rdma/rdma_backend.c
+++ b/hw/rdma/rdma_backend.c
@@ -64,9 +64,9 @@ static inline void complete_work(enum ibv_wc_status status, uint32_t vendor_err,
comp_handler(ctx, &wc);
}
-static void rdma_poll_cq(RdmaDeviceResources *rdma_dev_res, struct ibv_cq *ibcq)
+static int rdma_poll_cq(RdmaDeviceResources *rdma_dev_res, struct ibv_cq *ibcq)
{
- int i, ne;
+ int i, ne, total_ne = 0;
BackendCtx *bctx;
struct ibv_wc wc[2];
@@ -89,12 +89,18 @@ static void rdma_poll_cq(RdmaDeviceResources *rdma_dev_res, struct ibv_cq *ibcq)
rdma_rm_dealloc_cqe_ctx(rdma_dev_res, wc[i].wr_id);
g_free(bctx);
}
+ total_ne += ne;
} while (ne > 0);
+ atomic_sub(&rdma_dev_res->stats.missing_cqe, total_ne);
qemu_mutex_unlock(&rdma_dev_res->lock);
if (ne < 0) {
rdma_error_report("ibv_poll_cq fail, rc=%d, errno=%d", ne, errno);
}
+
+ rdma_dev_res->stats.completions += total_ne;
+
+ return total_ne;
}
static void *comp_handler_thread(void *arg)
@@ -122,6 +128,9 @@ static void *comp_handler_thread(void *arg)
while (backend_dev->comp_thread.run) {
do {
rc = qemu_poll_ns(pfds, 1, THR_POLL_TO * (int64_t)SCALE_MS);
+ if (!rc) {
+ backend_dev->rdma_dev_res->stats.poll_cq_ppoll_to++;
+ }
} while (!rc && backend_dev->comp_thread.run);
if (backend_dev->comp_thread.run) {
@@ -138,6 +147,7 @@ static void *comp_handler_thread(void *arg)
errno);
}
+ backend_dev->rdma_dev_res->stats.poll_cq_from_bk++;
rdma_poll_cq(backend_dev->rdma_dev_res, ev_cq);
ibv_ack_cq_events(ev_cq, 1);
@@ -271,7 +281,13 @@ int rdma_backend_query_port(RdmaBackendDev *backend_dev,
void rdma_backend_poll_cq(RdmaDeviceResources *rdma_dev_res, RdmaBackendCQ *cq)
{
- rdma_poll_cq(rdma_dev_res, cq->ibcq);
+ int polled;
+
+ rdma_dev_res->stats.poll_cq_from_guest++;
+ polled = rdma_poll_cq(rdma_dev_res, cq->ibcq);
+ if (!polled) {
+ rdma_dev_res->stats.poll_cq_from_guest_empty++;
+ }
}
static GHashTable *ah_hash;
@@ -333,7 +349,7 @@ static void ah_cache_init(void)
static int build_host_sge_array(RdmaDeviceResources *rdma_dev_res,
struct ibv_sge *dsge, struct ibv_sge *ssge,
- uint8_t num_sge)
+ uint8_t num_sge, uint64_t *total_length)
{
RdmaRmMR *mr;
int ssge_idx;
@@ -349,6 +365,8 @@ static int build_host_sge_array(RdmaDeviceResources *rdma_dev_res,
dsge->length = ssge[ssge_idx].length;
dsge->lkey = rdma_backend_mr_lkey(&mr->backend_mr);
+ *total_length += dsge->length;
+
dsge++;
}
@@ -445,8 +463,10 @@ void rdma_backend_post_send(RdmaBackendDev *backend_dev,
rc = mad_send(backend_dev, sgid_idx, sgid, sge, num_sge);
if (rc) {
complete_work(IBV_WC_GENERAL_ERR, VENDOR_ERR_MAD_SEND, ctx);
+ backend_dev->rdma_dev_res->stats.mad_tx_err++;
} else {
complete_work(IBV_WC_SUCCESS, 0, ctx);
+ backend_dev->rdma_dev_res->stats.mad_tx++;
}
}
return;
@@ -458,20 +478,21 @@ void rdma_backend_post_send(RdmaBackendDev *backend_dev,
rc = rdma_rm_alloc_cqe_ctx(backend_dev->rdma_dev_res, &bctx_id, bctx);
if (unlikely(rc)) {
complete_work(IBV_WC_GENERAL_ERR, VENDOR_ERR_NOMEM, ctx);
- goto out_free_bctx;
+ goto err_free_bctx;
}
- rc = build_host_sge_array(backend_dev->rdma_dev_res, new_sge, sge, num_sge);
+ rc = build_host_sge_array(backend_dev->rdma_dev_res, new_sge, sge, num_sge,
+ &backend_dev->rdma_dev_res->stats.tx_len);
if (rc) {
complete_work(IBV_WC_GENERAL_ERR, rc, ctx);
- goto out_dealloc_cqe_ctx;
+ goto err_dealloc_cqe_ctx;
}
if (qp_type == IBV_QPT_UD) {
wr.wr.ud.ah = create_ah(backend_dev, qp->ibpd, sgid_idx, dgid);
if (!wr.wr.ud.ah) {
complete_work(IBV_WC_GENERAL_ERR, VENDOR_ERR_FAIL_BACKEND, ctx);
- goto out_dealloc_cqe_ctx;
+ goto err_dealloc_cqe_ctx;
}
wr.wr.ud.remote_qpn = dqpn;
wr.wr.ud.remote_qkey = dqkey;
@@ -488,15 +509,19 @@ void rdma_backend_post_send(RdmaBackendDev *backend_dev,
rdma_error_report("ibv_post_send fail, qpn=0x%x, rc=%d, errno=%d",
qp->ibqp->qp_num, rc, errno);
complete_work(IBV_WC_GENERAL_ERR, VENDOR_ERR_FAIL_BACKEND, ctx);
- goto out_dealloc_cqe_ctx;
+ goto err_dealloc_cqe_ctx;
}
+ atomic_inc(&backend_dev->rdma_dev_res->stats.missing_cqe);
+ backend_dev->rdma_dev_res->stats.tx++;
+
return;
-out_dealloc_cqe_ctx:
+err_dealloc_cqe_ctx:
+ backend_dev->rdma_dev_res->stats.tx_err++;
rdma_rm_dealloc_cqe_ctx(backend_dev->rdma_dev_res, bctx_id);
-out_free_bctx:
+err_free_bctx:
g_free(bctx);
}
@@ -554,6 +579,9 @@ void rdma_backend_post_recv(RdmaBackendDev *backend_dev,
rc = save_mad_recv_buffer(backend_dev, sge, num_sge, ctx);
if (rc) {
complete_work(IBV_WC_GENERAL_ERR, rc, ctx);
+ rdma_dev_res->stats.mad_rx_bufs_err++;
+ } else {
+ rdma_dev_res->stats.mad_rx_bufs++;
}
}
return;
@@ -565,13 +593,14 @@ void rdma_backend_post_recv(RdmaBackendDev *backend_dev,
rc = rdma_rm_alloc_cqe_ctx(rdma_dev_res, &bctx_id, bctx);
if (unlikely(rc)) {
complete_work(IBV_WC_GENERAL_ERR, VENDOR_ERR_NOMEM, ctx);
- goto out_free_bctx;
+ goto err_free_bctx;
}
- rc = build_host_sge_array(rdma_dev_res, new_sge, sge, num_sge);
+ rc = build_host_sge_array(rdma_dev_res, new_sge, sge, num_sge,
+ &backend_dev->rdma_dev_res->stats.rx_bufs_len);
if (rc) {
complete_work(IBV_WC_GENERAL_ERR, rc, ctx);
- goto out_dealloc_cqe_ctx;
+ goto err_dealloc_cqe_ctx;
}
wr.num_sge = num_sge;
@@ -582,15 +611,19 @@ void rdma_backend_post_recv(RdmaBackendDev *backend_dev,
rdma_error_report("ibv_post_recv fail, qpn=0x%x, rc=%d, errno=%d",
qp->ibqp->qp_num, rc, errno);
complete_work(IBV_WC_GENERAL_ERR, VENDOR_ERR_FAIL_BACKEND, ctx);
- goto out_dealloc_cqe_ctx;
+ goto err_dealloc_cqe_ctx;
}
+ atomic_inc(&backend_dev->rdma_dev_res->stats.missing_cqe);
+ rdma_dev_res->stats.rx_bufs++;
+
return;
-out_dealloc_cqe_ctx:
+err_dealloc_cqe_ctx:
+ backend_dev->rdma_dev_res->stats.rx_bufs_err++;
rdma_rm_dealloc_cqe_ctx(rdma_dev_res, bctx_id);
-out_free_bctx:
+err_free_bctx:
g_free(bctx);
}
@@ -929,12 +962,14 @@ static void process_incoming_mad_req(RdmaBackendDev *backend_dev,
bctx = rdma_rm_get_cqe_ctx(backend_dev->rdma_dev_res, cqe_ctx_id);
if (unlikely(!bctx)) {
rdma_error_report("No matching ctx for req %ld", cqe_ctx_id);
+ backend_dev->rdma_dev_res->stats.mad_rx_err++;
return;
}
mad = rdma_pci_dma_map(backend_dev->dev, bctx->sge.addr,
bctx->sge.length);
if (!mad || bctx->sge.length < msg->umad_len + MAD_HDR_SIZE) {
+ backend_dev->rdma_dev_res->stats.mad_rx_err++;
complete_work(IBV_WC_GENERAL_ERR, VENDOR_ERR_INV_MAD_BUFF,
bctx->up_ctx);
} else {
@@ -949,6 +984,7 @@ static void process_incoming_mad_req(RdmaBackendDev *backend_dev,
wc.byte_len = msg->umad_len;
wc.status = IBV_WC_SUCCESS;
wc.wc_flags = IBV_WC_GRH;
+ backend_dev->rdma_dev_res->stats.mad_rx++;
comp_handler(bctx->up_ctx, &wc);
}
diff --git a/hw/rdma/rdma_hmp.c b/hw/rdma/rdma_hmp.c
new file mode 100644
index 0000000000..c5814473c5
--- /dev/null
+++ b/hw/rdma/rdma_hmp.c
@@ -0,0 +1,30 @@
+/*
+ * RDMA device: Human Monitor interface
+ *
+ * Copyright (C) 2018 Oracle
+ * Copyright (C) 2018 Red Hat Inc
+ *
+ * Authors:
+ * Yuval Shaia <yuval.shaia@oracle.com>
+ *
+ * This work is licensed under the terms of the GNU GPL, version 2 or later.
+ * See the COPYING file in the top-level directory.
+ *
+ */
+
+#include "qemu/osdep.h"
+#include "hw/rdma/rdma_hmp.h"
+#include "qemu/module.h"
+
+static const TypeInfo rdma_hmp_info = {
+ .name = TYPE_RDMA_STATS_PROVIDER,
+ .parent = TYPE_INTERFACE,
+ .class_size = sizeof(RdmaStatsProviderClass),
+};
+
+static void rdma_hmp_register_types(void)
+{
+ type_register_static(&rdma_hmp_info);
+}
+
+type_init(rdma_hmp_register_types)
diff --git a/hw/rdma/rdma_rm.c b/hw/rdma/rdma_rm.c
index 14580ca379..e019de1a14 100644
--- a/hw/rdma/rdma_rm.c
+++ b/hw/rdma/rdma_rm.c
@@ -16,6 +16,7 @@
#include "qemu/osdep.h"
#include "qapi/error.h"
#include "cpu.h"
+#include "monitor/monitor.h"
#include "trace.h"
#include "rdma_utils.h"
@@ -26,6 +27,58 @@
#define PG_DIR_SZ { TARGET_PAGE_SIZE / sizeof(__u64) }
#define PG_TBL_SZ { TARGET_PAGE_SIZE / sizeof(__u64) }
+void rdma_dump_device_counters(Monitor *mon, RdmaDeviceResources *dev_res)
+{
+ monitor_printf(mon, "\ttx : %" PRId64 "\n",
+ dev_res->stats.tx);
+ monitor_printf(mon, "\ttx_len : %" PRId64 "\n",
+ dev_res->stats.tx_len);
+ monitor_printf(mon, "\ttx_err : %" PRId64 "\n",
+ dev_res->stats.tx_err);
+ monitor_printf(mon, "\trx_bufs : %" PRId64 "\n",
+ dev_res->stats.rx_bufs);
+ monitor_printf(mon, "\trx_bufs_len : %" PRId64 "\n",
+ dev_res->stats.rx_bufs_len);
+ monitor_printf(mon, "\trx_bufs_err : %" PRId64 "\n",
+ dev_res->stats.rx_bufs_err);
+ monitor_printf(mon, "\tcomps : %" PRId64 "\n",
+ dev_res->stats.completions);
+ monitor_printf(mon, "\tmissing_comps : %" PRId32 "\n",
+ dev_res->stats.missing_cqe);
+ monitor_printf(mon, "\tpoll_cq (bk) : %" PRId64 "\n",
+ dev_res->stats.poll_cq_from_bk);
+ monitor_printf(mon, "\tpoll_cq_ppoll_to : %" PRId64 "\n",
+ dev_res->stats.poll_cq_ppoll_to);
+ monitor_printf(mon, "\tpoll_cq (fe) : %" PRId64 "\n",
+ dev_res->stats.poll_cq_from_guest);
+ monitor_printf(mon, "\tpoll_cq_empty : %" PRId64 "\n",
+ dev_res->stats.poll_cq_from_guest_empty);
+ monitor_printf(mon, "\tmad_tx : %" PRId64 "\n",
+ dev_res->stats.mad_tx);
+ monitor_printf(mon, "\tmad_tx_err : %" PRId64 "\n",
+ dev_res->stats.mad_tx_err);
+ monitor_printf(mon, "\tmad_rx : %" PRId64 "\n",
+ dev_res->stats.mad_rx);
+ monitor_printf(mon, "\tmad_rx_err : %" PRId64 "\n",
+ dev_res->stats.mad_rx_err);
+ monitor_printf(mon, "\tmad_rx_bufs : %" PRId64 "\n",
+ dev_res->stats.mad_rx_bufs);
+ monitor_printf(mon, "\tmad_rx_bufs_err : %" PRId64 "\n",
+ dev_res->stats.mad_rx_bufs_err);
+ monitor_printf(mon, "\tPDs : %" PRId32 "\n",
+ dev_res->pd_tbl.used);
+ monitor_printf(mon, "\tMRs : %" PRId32 "\n",
+ dev_res->mr_tbl.used);
+ monitor_printf(mon, "\tUCs : %" PRId32 "\n",
+ dev_res->uc_tbl.used);
+ monitor_printf(mon, "\tQPs : %" PRId32 "\n",
+ dev_res->qp_tbl.used);
+ monitor_printf(mon, "\tCQs : %" PRId32 "\n",
+ dev_res->cq_tbl.used);
+ monitor_printf(mon, "\tCEQ_CTXs : %" PRId32 "\n",
+ dev_res->cqe_ctx_tbl.used);
+}
+
static inline void res_tbl_init(const char *name, RdmaRmResTbl *tbl,
uint32_t tbl_sz, uint32_t res_sz)
{
@@ -37,6 +90,7 @@ static inline void res_tbl_init(const char *name, RdmaRmResTbl *tbl,
tbl->bitmap = bitmap_new(tbl_sz);
tbl->tbl_sz = tbl_sz;
tbl->res_sz = res_sz;
+ tbl->used = 0;
qemu_mutex_init(&tbl->lock);
}
@@ -76,6 +130,8 @@ static inline void *rdma_res_tbl_alloc(RdmaRmResTbl *tbl, uint32_t *handle)
set_bit(*handle, tbl->bitmap);
+ tbl->used++;
+
qemu_mutex_unlock(&tbl->lock);
memset(tbl->tbl + *handle * tbl->res_sz, 0, tbl->res_sz);
@@ -93,6 +149,7 @@ static inline void rdma_res_tbl_dealloc(RdmaRmResTbl *tbl, uint32_t handle)
if (handle < tbl->tbl_sz) {
clear_bit(handle, tbl->bitmap);
+ tbl->used--;
}
qemu_mutex_unlock(&tbl->lock);
@@ -620,6 +677,9 @@ int rdma_rm_init(RdmaDeviceResources *dev_res, struct ibv_device_attr *dev_attr,
qemu_mutex_init(&dev_res->lock);
+ memset(&dev_res->stats, 0, sizeof(dev_res->stats));
+ atomic_set(&dev_res->stats.missing_cqe, 0);
+
return 0;
}
diff --git a/hw/rdma/rdma_rm.h b/hw/rdma/rdma_rm.h
index 9ec87d2667..a527d306f6 100644
--- a/hw/rdma/rdma_rm.h
+++ b/hw/rdma/rdma_rm.h
@@ -20,6 +20,7 @@
#include "rdma_backend_defs.h"
#include "rdma_rm_defs.h"
+void rdma_dump_device_counters(Monitor *mon, RdmaDeviceResources *dev_res);
int rdma_rm_init(RdmaDeviceResources *dev_res, struct ibv_device_attr *dev_attr,
Error **errp);
void rdma_rm_fini(RdmaDeviceResources *dev_res, RdmaBackendDev *backend_dev,
diff --git a/hw/rdma/rdma_rm_defs.h b/hw/rdma/rdma_rm_defs.h
index f0ee1f3072..4b8d704cfe 100644
--- a/hw/rdma/rdma_rm_defs.h
+++ b/hw/rdma/rdma_rm_defs.h
@@ -34,7 +34,9 @@
#define MAX_QP_INIT_RD_ATOM 16
#define MAX_AH 64
-#define MAX_RM_TBL_NAME 16
+#define MAX_RM_TBL_NAME 16
+#define MAX_CONSEQ_EMPTY_POLL_CQ 4096 /* considered as error above this */
+
typedef struct RdmaRmResTbl {
char name[MAX_RM_TBL_NAME];
QemuMutex lock;
@@ -42,6 +44,7 @@ typedef struct RdmaRmResTbl {
size_t tbl_sz;
size_t res_sz;
void *tbl;
+ uint32_t used; /* number of used entries in the table */
} RdmaRmResTbl;
typedef struct RdmaRmPD {
@@ -96,6 +99,27 @@ typedef struct RdmaRmPort {
enum ibv_port_state state;
} RdmaRmPort;
+typedef struct RdmaRmStats {
+ uint64_t tx;
+ uint64_t tx_len;
+ uint64_t tx_err;
+ uint64_t rx_bufs;
+ uint64_t rx_bufs_len;
+ uint64_t rx_bufs_err;
+ uint64_t completions;
+ uint64_t mad_tx;
+ uint64_t mad_tx_err;
+ uint64_t mad_rx;
+ uint64_t mad_rx_err;
+ uint64_t mad_rx_bufs;
+ uint64_t mad_rx_bufs_err;
+ uint64_t poll_cq_from_bk;
+ uint64_t poll_cq_from_guest;
+ uint64_t poll_cq_from_guest_empty;
+ uint64_t poll_cq_ppoll_to;
+ uint32_t missing_cqe;
+} RdmaRmStats;
+
typedef struct RdmaDeviceResources {
RdmaRmPort port;
RdmaRmResTbl pd_tbl;
@@ -106,6 +130,7 @@ typedef struct RdmaDeviceResources {
RdmaRmResTbl cqe_ctx_tbl;
GHashTable *qp_hash; /* Keeps mapping between real and emulated */
QemuMutex lock;
+ RdmaRmStats stats;
} RdmaDeviceResources;
#endif
diff --git a/hw/rdma/vmw/pvrdma.h b/hw/rdma/vmw/pvrdma.h
index 0879224957..167706ec2c 100644
--- a/hw/rdma/vmw/pvrdma.h
+++ b/hw/rdma/vmw/pvrdma.h
@@ -70,6 +70,10 @@ typedef struct DSRInfo {
PvrdmaRing cq;
} DSRInfo;
+typedef struct PVRDMADevStats {
+ uint64_t commands;
+} PVRDMADevStats;
+
typedef struct PVRDMADev {
PCIDevice parent_obj;
MemoryRegion msix;
@@ -89,6 +93,7 @@ typedef struct PVRDMADev {
CharBackend mad_chr;
VMXNET3State *func0;
Notifier shutdown_notifier;
+ PVRDMADevStats stats;
} PVRDMADev;
#define PVRDMA_DEV(dev) OBJECT_CHECK(PVRDMADev, (dev), PVRDMA_HW_NAME)
diff --git a/hw/rdma/vmw/pvrdma_main.c b/hw/rdma/vmw/pvrdma_main.c
index b6061f4b6e..659331ac93 100644
--- a/hw/rdma/vmw/pvrdma_main.c
+++ b/hw/rdma/vmw/pvrdma_main.c
@@ -25,6 +25,8 @@
#include "cpu.h"
#include "trace.h"
#include "sysemu/sysemu.h"
+#include "monitor/monitor.h"
+#include "hw/rdma/rdma_hmp.h"
#include "../rdma_rm.h"
#include "../rdma_backend.h"
@@ -55,6 +57,18 @@ static Property pvrdma_dev_properties[] = {
DEFINE_PROP_END_OF_LIST(),
};
+static void pvrdma_print_statistics(Monitor *mon, RdmaStatsProvider *obj)
+{
+ PVRDMADev *dev = PVRDMA_DEV(obj);
+ PCIDevice *pdev = PCI_DEVICE(dev);
+
+ monitor_printf(mon, "%s, %x.%x\n", pdev->name, PCI_SLOT(pdev->devfn),
+ PCI_FUNC(pdev->devfn));
+ monitor_printf(mon, "\tcommands : %" PRId64 "\n",
+ dev->stats.commands);
+ rdma_dump_device_counters(mon, &dev->rdma_dev_res);
+}
+
static void free_dev_ring(PCIDevice *pci_dev, PvrdmaRing *ring,
void *ring_state)
{
@@ -394,6 +408,7 @@ static void pvrdma_regs_write(void *opaque, hwaddr addr, uint64_t val,
if (val == 0) {
trace_pvrdma_regs_write(addr, val, "REQUEST", "");
pvrdma_exec_cmd(dev);
+ dev->stats.commands++;
}
break;
default:
@@ -612,6 +627,8 @@ static void pvrdma_realize(PCIDevice *pdev, Error **errp)
goto out;
}
+ memset(&dev->stats, 0, sizeof(dev->stats));
+
dev->shutdown_notifier.notify = pvrdma_shutdown_notifier;
qemu_register_shutdown_notifier(&dev->shutdown_notifier);
@@ -631,6 +648,7 @@ static void pvrdma_class_init(ObjectClass *klass, void *data)
{
DeviceClass *dc = DEVICE_CLASS(klass);
PCIDeviceClass *k = PCI_DEVICE_CLASS(klass);
+ RdmaStatsProviderClass *ir = RDMA_STATS_PROVIDER_CLASS(klass);
k->realize = pvrdma_realize;
k->exit = pvrdma_exit;
@@ -642,6 +660,8 @@ static void pvrdma_class_init(ObjectClass *klass, void *data)
dc->desc = "RDMA Device";
dc->props = pvrdma_dev_properties;
set_bit(DEVICE_CATEGORY_NETWORK, dc->categories);
+
+ ir->print_statistics = pvrdma_print_statistics;
}
static const TypeInfo pvrdma_info = {
@@ -651,6 +671,7 @@ static const TypeInfo pvrdma_info = {
.class_init = pvrdma_class_init,
.interfaces = (InterfaceInfo[]) {
{ INTERFACE_CONVENTIONAL_PCI_DEVICE },
+ { TYPE_RDMA_STATS_PROVIDER },
{ }
}
};
diff --git a/include/hw/rdma/rdma_hmp.h b/include/hw/rdma/rdma_hmp.h
new file mode 100644
index 0000000000..dd23f2bc84
--- /dev/null
+++ b/include/hw/rdma/rdma_hmp.h
@@ -0,0 +1,40 @@
+/*
+ * RDMA device: Human Monitor interface
+ *
+ * Copyright (C) 2019 Oracle
+ * Copyright (C) 2019 Red Hat Inc
+ *
+ * Authors:
+ * Yuval Shaia <yuval.shaia@oracle.com>
+ *
+ * This work is licensed under the terms of the GNU GPL, version 2 or later.
+ * See the COPYING file in the top-level directory.
+ *
+ */
+
+#ifndef RDMA_HMP_H
+#define RDMA_HMP_H
+
+#include "qom/object.h"
+
+#define TYPE_RDMA_STATS_PROVIDER "rdma"
+
+#define RDMA_STATS_PROVIDER_CLASS(klass) \
+ OBJECT_CLASS_CHECK(RdmaStatsProviderClass, (klass), \
+ TYPE_RDMA_STATS_PROVIDER)
+#define RDMA_STATS_PROVIDER_GET_CLASS(obj) \
+ OBJECT_GET_CLASS(RdmaStatsProviderClass, (obj), \
+ TYPE_RDMA_STATS_PROVIDER)
+#define RDMA_STATS_PROVIDER(obj) \
+ INTERFACE_CHECK(RdmaStatsProvider, (obj), \
+ TYPE_RDMA_STATS_PROVIDER)
+
+typedef struct RdmaStatsProvider RdmaStatsProvider;
+
+typedef struct RdmaStatsProviderClass {
+ InterfaceClass parent;
+
+ void (*print_statistics)(Monitor *mon, RdmaStatsProvider *obj);
+} RdmaStatsProviderClass;
+
+#endif
--
2.17.2
^ permalink raw reply related [flat|nested] 17+ messages in thread
* [Qemu-devel] [PATCH v4 5/9] hw/rdma: Free all MAD receive buffers when device is closed
2019-03-03 20:33 [Qemu-devel] [PATCH v4 0/9] Misc fixes to pvrdma device Yuval Shaia
` (3 preceding siblings ...)
2019-03-03 20:33 ` [Qemu-devel] [PATCH v4 4/9] {hmp, hw/pvrdma}: Expose device internals via monitor interface Yuval Shaia
@ 2019-03-03 20:33 ` Yuval Shaia
2019-03-03 20:33 ` [Qemu-devel] [PATCH v4 6/9] hw/rdma: Free all receive buffers when QP is destroyed Yuval Shaia
` (3 subsequent siblings)
8 siblings, 0 replies; 17+ messages in thread
From: Yuval Shaia @ 2019-03-03 20:33 UTC (permalink / raw)
To: dgilbert, yuval.shaia, marcel.apfelbaum, armbru, qemu-devel
When device is going down free all saved MAD buffers.
Signed-off-by: Yuval Shaia <yuval.shaia@oracle.com>
Reviewed-by: Marcel Apfelbaum <marcel.apfelbaum@gmail.com>
---
hw/rdma/rdma_backend.c | 34 +++++++++++++++++++++++++++++++++-
hw/rdma/vmw/pvrdma_main.c | 2 ++
2 files changed, 35 insertions(+), 1 deletion(-)
diff --git a/hw/rdma/rdma_backend.c b/hw/rdma/rdma_backend.c
index bc2fefcf93..a65f5737e4 100644
--- a/hw/rdma/rdma_backend.c
+++ b/hw/rdma/rdma_backend.c
@@ -64,6 +64,33 @@ static inline void complete_work(enum ibv_wc_status status, uint32_t vendor_err,
comp_handler(ctx, &wc);
}
+static void free_cqe_ctx(gpointer data, gpointer user_data)
+{
+ BackendCtx *bctx;
+ RdmaDeviceResources *rdma_dev_res = user_data;
+ unsigned long cqe_ctx_id = GPOINTER_TO_INT(data);
+
+ bctx = rdma_rm_get_cqe_ctx(rdma_dev_res, cqe_ctx_id);
+ if (bctx) {
+ rdma_rm_dealloc_cqe_ctx(rdma_dev_res, cqe_ctx_id);
+ }
+ g_free(bctx);
+}
+
+static void clean_recv_mads(RdmaBackendDev *backend_dev)
+{
+ unsigned long cqe_ctx_id;
+
+ do {
+ cqe_ctx_id = rdma_protected_qlist_pop_int64(&backend_dev->
+ recv_mads_list);
+ if (cqe_ctx_id != -ENOENT) {
+ free_cqe_ctx(GINT_TO_POINTER(cqe_ctx_id),
+ backend_dev->rdma_dev_res);
+ }
+ } while (cqe_ctx_id != -ENOENT);
+}
+
static int rdma_poll_cq(RdmaDeviceResources *rdma_dev_res, struct ibv_cq *ibcq)
{
int i, ne, total_ne = 0;
@@ -1037,6 +1064,11 @@ static int mad_init(RdmaBackendDev *backend_dev, CharBackend *mad_chr_be)
return 0;
}
+static void mad_stop(RdmaBackendDev *backend_dev)
+{
+ clean_recv_mads(backend_dev);
+}
+
static void mad_fini(RdmaBackendDev *backend_dev)
{
disable_rdmacm_mux_async(backend_dev);
@@ -1224,12 +1256,12 @@ void rdma_backend_start(RdmaBackendDev *backend_dev)
void rdma_backend_stop(RdmaBackendDev *backend_dev)
{
+ mad_stop(backend_dev);
stop_backend_thread(&backend_dev->comp_thread);
}
void rdma_backend_fini(RdmaBackendDev *backend_dev)
{
- rdma_backend_stop(backend_dev);
mad_fini(backend_dev);
g_hash_table_destroy(ah_hash);
ibv_destroy_comp_channel(backend_dev->channel);
diff --git a/hw/rdma/vmw/pvrdma_main.c b/hw/rdma/vmw/pvrdma_main.c
index 659331ac93..b795f80666 100644
--- a/hw/rdma/vmw/pvrdma_main.c
+++ b/hw/rdma/vmw/pvrdma_main.c
@@ -305,6 +305,8 @@ static void pvrdma_fini(PCIDevice *pdev)
pvrdma_qp_ops_fini();
+ rdma_backend_stop(&dev->backend_dev);
+
rdma_rm_fini(&dev->rdma_dev_res, &dev->backend_dev,
dev->backend_eth_device_name);
--
2.17.2
^ permalink raw reply related [flat|nested] 17+ messages in thread
* [Qemu-devel] [PATCH v4 6/9] hw/rdma: Free all receive buffers when QP is destroyed
2019-03-03 20:33 [Qemu-devel] [PATCH v4 0/9] Misc fixes to pvrdma device Yuval Shaia
` (4 preceding siblings ...)
2019-03-03 20:33 ` [Qemu-devel] [PATCH v4 5/9] hw/rdma: Free all MAD receive buffers when device is closed Yuval Shaia
@ 2019-03-03 20:33 ` Yuval Shaia
2019-03-03 20:33 ` [Qemu-devel] [PATCH v4 7/9] hw/pvrdma: Delete unneeded function argument Yuval Shaia
` (2 subsequent siblings)
8 siblings, 0 replies; 17+ messages in thread
From: Yuval Shaia @ 2019-03-03 20:33 UTC (permalink / raw)
To: dgilbert, yuval.shaia, marcel.apfelbaum, armbru, qemu-devel
When QP is destroyed the backend QP is destroyed as well. This ensures
we clean all received buffer we posted to it.
However, a contexts of these buffers are still remain in the device.
Fix it by maintaining a list of buffer's context and free them when QP
is destroyed.
Signed-off-by: Yuval Shaia <yuval.shaia@oracle.com>
Reviewed-by: Marcel Apfelbaum <marcel.apfelbaum@gmail.com>
---
hw/rdma/rdma_backend.c | 26 ++++++++++++++++++++------
hw/rdma/rdma_backend.h | 2 +-
hw/rdma/rdma_backend_defs.h | 2 +-
hw/rdma/rdma_rm.c | 2 +-
hw/rdma/rdma_utils.c | 29 +++++++++++++++++++++++++++++
hw/rdma/rdma_utils.h | 11 +++++++++++
6 files changed, 63 insertions(+), 9 deletions(-)
diff --git a/hw/rdma/rdma_backend.c b/hw/rdma/rdma_backend.c
index a65f5737e4..d511ca282b 100644
--- a/hw/rdma/rdma_backend.c
+++ b/hw/rdma/rdma_backend.c
@@ -39,6 +39,7 @@
typedef struct BackendCtx {
void *up_ctx;
struct ibv_sge sge; /* Used to save MAD recv buffer */
+ RdmaBackendQP *backend_qp; /* To maintain recv buffers */
} BackendCtx;
struct backend_umad {
@@ -73,6 +74,7 @@ static void free_cqe_ctx(gpointer data, gpointer user_data)
bctx = rdma_rm_get_cqe_ctx(rdma_dev_res, cqe_ctx_id);
if (bctx) {
rdma_rm_dealloc_cqe_ctx(rdma_dev_res, cqe_ctx_id);
+ atomic_dec(&rdma_dev_res->stats.missing_cqe);
}
g_free(bctx);
}
@@ -85,13 +87,15 @@ static void clean_recv_mads(RdmaBackendDev *backend_dev)
cqe_ctx_id = rdma_protected_qlist_pop_int64(&backend_dev->
recv_mads_list);
if (cqe_ctx_id != -ENOENT) {
+ atomic_inc(&backend_dev->rdma_dev_res->stats.missing_cqe);
free_cqe_ctx(GINT_TO_POINTER(cqe_ctx_id),
backend_dev->rdma_dev_res);
}
} while (cqe_ctx_id != -ENOENT);
}
-static int rdma_poll_cq(RdmaDeviceResources *rdma_dev_res, struct ibv_cq *ibcq)
+static int rdma_poll_cq(RdmaBackendDev *backend_dev,
+ RdmaDeviceResources *rdma_dev_res, struct ibv_cq *ibcq)
{
int i, ne, total_ne = 0;
BackendCtx *bctx;
@@ -113,6 +117,8 @@ static int rdma_poll_cq(RdmaDeviceResources *rdma_dev_res, struct ibv_cq *ibcq)
comp_handler(bctx->up_ctx, &wc[i]);
+ rdma_protected_gslist_remove_int32(&bctx->backend_qp->cqe_ctx_list,
+ wc[i].wr_id);
rdma_rm_dealloc_cqe_ctx(rdma_dev_res, wc[i].wr_id);
g_free(bctx);
}
@@ -175,14 +181,12 @@ static void *comp_handler_thread(void *arg)
}
backend_dev->rdma_dev_res->stats.poll_cq_from_bk++;
- rdma_poll_cq(backend_dev->rdma_dev_res, ev_cq);
+ rdma_poll_cq(backend_dev, backend_dev->rdma_dev_res, ev_cq);
ibv_ack_cq_events(ev_cq, 1);
}
}
- /* TODO: Post cqe for all remaining buffs that were posted */
-
backend_dev->comp_thread.is_running = false;
qemu_thread_exit(0);
@@ -311,7 +315,7 @@ void rdma_backend_poll_cq(RdmaDeviceResources *rdma_dev_res, RdmaBackendCQ *cq)
int polled;
rdma_dev_res->stats.poll_cq_from_guest++;
- polled = rdma_poll_cq(rdma_dev_res, cq->ibcq);
+ polled = rdma_poll_cq(cq->backend_dev, rdma_dev_res, cq->ibcq);
if (!polled) {
rdma_dev_res->stats.poll_cq_from_guest_empty++;
}
@@ -501,6 +505,7 @@ void rdma_backend_post_send(RdmaBackendDev *backend_dev,
bctx = g_malloc0(sizeof(*bctx));
bctx->up_ctx = ctx;
+ bctx->backend_qp = qp;
rc = rdma_rm_alloc_cqe_ctx(backend_dev->rdma_dev_res, &bctx_id, bctx);
if (unlikely(rc)) {
@@ -508,6 +513,8 @@ void rdma_backend_post_send(RdmaBackendDev *backend_dev,
goto err_free_bctx;
}
+ rdma_protected_gslist_append_int32(&qp->cqe_ctx_list, bctx_id);
+
rc = build_host_sge_array(backend_dev->rdma_dev_res, new_sge, sge, num_sge,
&backend_dev->rdma_dev_res->stats.tx_len);
if (rc) {
@@ -616,6 +623,7 @@ void rdma_backend_post_recv(RdmaBackendDev *backend_dev,
bctx = g_malloc0(sizeof(*bctx));
bctx->up_ctx = ctx;
+ bctx->backend_qp = qp;
rc = rdma_rm_alloc_cqe_ctx(rdma_dev_res, &bctx_id, bctx);
if (unlikely(rc)) {
@@ -623,6 +631,8 @@ void rdma_backend_post_recv(RdmaBackendDev *backend_dev,
goto err_free_bctx;
}
+ rdma_protected_gslist_append_int32(&qp->cqe_ctx_list, bctx_id);
+
rc = build_host_sge_array(rdma_dev_res, new_sge, sge, num_sge,
&backend_dev->rdma_dev_res->stats.rx_bufs_len);
if (rc) {
@@ -762,6 +772,8 @@ int rdma_backend_create_qp(RdmaBackendQP *qp, uint8_t qp_type,
return -EIO;
}
+ rdma_protected_gslist_init(&qp->cqe_ctx_list);
+
qp->ibpd = pd->ibpd;
/* TODO: Query QP to get max_inline_data and save it to be used in send */
@@ -919,11 +931,13 @@ int rdma_backend_query_qp(RdmaBackendQP *qp, struct ibv_qp_attr *attr,
return ibv_query_qp(qp->ibqp, attr, attr_mask, init_attr);
}
-void rdma_backend_destroy_qp(RdmaBackendQP *qp)
+void rdma_backend_destroy_qp(RdmaBackendQP *qp, RdmaDeviceResources *dev_res)
{
if (qp->ibqp) {
ibv_destroy_qp(qp->ibqp);
}
+ g_slist_foreach(qp->cqe_ctx_list.list, free_cqe_ctx, dev_res);
+ rdma_protected_gslist_destroy(&qp->cqe_ctx_list);
}
#define CHK_ATTR(req, dev, member, fmt) ({ \
diff --git a/hw/rdma/rdma_backend.h b/hw/rdma/rdma_backend.h
index 5114c90e67..cb5efa2a3a 100644
--- a/hw/rdma/rdma_backend.h
+++ b/hw/rdma/rdma_backend.h
@@ -102,7 +102,7 @@ int rdma_backend_qp_state_rts(RdmaBackendQP *qp, uint8_t qp_type,
uint32_t sq_psn, uint32_t qkey, bool use_qkey);
int rdma_backend_query_qp(RdmaBackendQP *qp, struct ibv_qp_attr *attr,
int attr_mask, struct ibv_qp_init_attr *init_attr);
-void rdma_backend_destroy_qp(RdmaBackendQP *qp);
+void rdma_backend_destroy_qp(RdmaBackendQP *qp, RdmaDeviceResources *dev_res);
void rdma_backend_post_send(RdmaBackendDev *backend_dev,
RdmaBackendQP *qp, uint8_t qp_type,
diff --git a/hw/rdma/rdma_backend_defs.h b/hw/rdma/rdma_backend_defs.h
index a8c15b09ab..817153dc8c 100644
--- a/hw/rdma/rdma_backend_defs.h
+++ b/hw/rdma/rdma_backend_defs.h
@@ -26,7 +26,6 @@ typedef struct RdmaDeviceResources RdmaDeviceResources;
typedef struct RdmaBackendThread {
QemuThread thread;
- QemuMutex mutex;
bool run; /* Set by thread manager to let thread know it should exit */
bool is_running; /* Set by the thread to report its status */
} RdmaBackendThread;
@@ -66,6 +65,7 @@ typedef struct RdmaBackendQP {
struct ibv_pd *ibpd;
struct ibv_qp *ibqp;
uint8_t sgid_idx;
+ RdmaProtectedGSList cqe_ctx_list;
} RdmaBackendQP;
#endif
diff --git a/hw/rdma/rdma_rm.c b/hw/rdma/rdma_rm.c
index e019de1a14..4c9fe1591e 100644
--- a/hw/rdma/rdma_rm.c
+++ b/hw/rdma/rdma_rm.c
@@ -537,7 +537,7 @@ void rdma_rm_dealloc_qp(RdmaDeviceResources *dev_res, uint32_t qp_handle)
return;
}
- rdma_backend_destroy_qp(&qp->backend_qp);
+ rdma_backend_destroy_qp(&qp->backend_qp, dev_res);
rdma_res_tbl_dealloc(&dev_res->qp_tbl, qp->qpn);
}
diff --git a/hw/rdma/rdma_utils.c b/hw/rdma/rdma_utils.c
index 0a8abe572d..73f279104c 100644
--- a/hw/rdma/rdma_utils.c
+++ b/hw/rdma/rdma_utils.c
@@ -90,3 +90,32 @@ int64_t rdma_protected_qlist_pop_int64(RdmaProtectedQList *list)
return qnum_get_uint(qobject_to(QNum, obj));
}
+
+void rdma_protected_gslist_init(RdmaProtectedGSList *list)
+{
+ qemu_mutex_init(&list->lock);
+}
+
+void rdma_protected_gslist_destroy(RdmaProtectedGSList *list)
+{
+ if (list->list) {
+ g_slist_free(list->list);
+ list->list = NULL;
+ }
+}
+
+void rdma_protected_gslist_append_int32(RdmaProtectedGSList *list,
+ int32_t value)
+{
+ qemu_mutex_lock(&list->lock);
+ list->list = g_slist_prepend(list->list, GINT_TO_POINTER(value));
+ qemu_mutex_unlock(&list->lock);
+}
+
+void rdma_protected_gslist_remove_int32(RdmaProtectedGSList *list,
+ int32_t value)
+{
+ qemu_mutex_lock(&list->lock);
+ list->list = g_slist_remove(list->list, GINT_TO_POINTER(value));
+ qemu_mutex_unlock(&list->lock);
+}
diff --git a/hw/rdma/rdma_utils.h b/hw/rdma/rdma_utils.h
index a8bf1d4fec..2d42249691 100644
--- a/hw/rdma/rdma_utils.h
+++ b/hw/rdma/rdma_utils.h
@@ -34,12 +34,23 @@ typedef struct RdmaProtectedQList {
QList *list;
} RdmaProtectedQList;
+typedef struct RdmaProtectedGSList {
+ QemuMutex lock;
+ GSList *list;
+} RdmaProtectedGSList;
+
void *rdma_pci_dma_map(PCIDevice *dev, dma_addr_t addr, dma_addr_t plen);
void rdma_pci_dma_unmap(PCIDevice *dev, void *buffer, dma_addr_t len);
void rdma_protected_qlist_init(RdmaProtectedQList *list);
void rdma_protected_qlist_destroy(RdmaProtectedQList *list);
void rdma_protected_qlist_append_int64(RdmaProtectedQList *list, int64_t value);
int64_t rdma_protected_qlist_pop_int64(RdmaProtectedQList *list);
+void rdma_protected_gslist_init(RdmaProtectedGSList *list);
+void rdma_protected_gslist_destroy(RdmaProtectedGSList *list);
+void rdma_protected_gslist_append_int32(RdmaProtectedGSList *list,
+ int32_t value);
+void rdma_protected_gslist_remove_int32(RdmaProtectedGSList *list,
+ int32_t value);
static inline void addrconf_addr_eui48(uint8_t *eui, const char *addr)
{
--
2.17.2
^ permalink raw reply related [flat|nested] 17+ messages in thread
* [Qemu-devel] [PATCH v4 7/9] hw/pvrdma: Delete unneeded function argument
2019-03-03 20:33 [Qemu-devel] [PATCH v4 0/9] Misc fixes to pvrdma device Yuval Shaia
` (5 preceding siblings ...)
2019-03-03 20:33 ` [Qemu-devel] [PATCH v4 6/9] hw/rdma: Free all receive buffers when QP is destroyed Yuval Shaia
@ 2019-03-03 20:33 ` Yuval Shaia
2019-03-03 20:33 ` [Qemu-devel] [PATCH v4 8/9] hw/pvrdma: Delete pvrdma_exit function Yuval Shaia
2019-03-03 20:33 ` [Qemu-devel] [PATCH v4 9/9] hw/pvrdma: Unregister from shutdown notifier when device goes down Yuval Shaia
8 siblings, 0 replies; 17+ messages in thread
From: Yuval Shaia @ 2019-03-03 20:33 UTC (permalink / raw)
To: dgilbert, yuval.shaia, marcel.apfelbaum, armbru, qemu-devel
The function's argument rdma_dev_res is not needed as it is stored in
the backend_dev object at init.
Signed-off-by: Yuval Shaia <yuval.shaia@oracle.com>
Reviewed-by: Marcel Apfelbaum <marcel.apfelbaum@gmail.com>
---
hw/rdma/rdma_backend.c | 13 ++++++-------
hw/rdma/rdma_backend.h | 1 -
hw/rdma/vmw/pvrdma_qp_ops.c | 3 +--
3 files changed, 7 insertions(+), 10 deletions(-)
diff --git a/hw/rdma/rdma_backend.c b/hw/rdma/rdma_backend.c
index d511ca282b..66185bd487 100644
--- a/hw/rdma/rdma_backend.c
+++ b/hw/rdma/rdma_backend.c
@@ -594,7 +594,6 @@ static unsigned int save_mad_recv_buffer(RdmaBackendDev *backend_dev,
}
void rdma_backend_post_recv(RdmaBackendDev *backend_dev,
- RdmaDeviceResources *rdma_dev_res,
RdmaBackendQP *qp, uint8_t qp_type,
struct ibv_sge *sge, uint32_t num_sge, void *ctx)
{
@@ -613,9 +612,9 @@ void rdma_backend_post_recv(RdmaBackendDev *backend_dev,
rc = save_mad_recv_buffer(backend_dev, sge, num_sge, ctx);
if (rc) {
complete_work(IBV_WC_GENERAL_ERR, rc, ctx);
- rdma_dev_res->stats.mad_rx_bufs_err++;
+ backend_dev->rdma_dev_res->stats.mad_rx_bufs_err++;
} else {
- rdma_dev_res->stats.mad_rx_bufs++;
+ backend_dev->rdma_dev_res->stats.mad_rx_bufs++;
}
}
return;
@@ -625,7 +624,7 @@ void rdma_backend_post_recv(RdmaBackendDev *backend_dev,
bctx->up_ctx = ctx;
bctx->backend_qp = qp;
- rc = rdma_rm_alloc_cqe_ctx(rdma_dev_res, &bctx_id, bctx);
+ rc = rdma_rm_alloc_cqe_ctx(backend_dev->rdma_dev_res, &bctx_id, bctx);
if (unlikely(rc)) {
complete_work(IBV_WC_GENERAL_ERR, VENDOR_ERR_NOMEM, ctx);
goto err_free_bctx;
@@ -633,7 +632,7 @@ void rdma_backend_post_recv(RdmaBackendDev *backend_dev,
rdma_protected_gslist_append_int32(&qp->cqe_ctx_list, bctx_id);
- rc = build_host_sge_array(rdma_dev_res, new_sge, sge, num_sge,
+ rc = build_host_sge_array(backend_dev->rdma_dev_res, new_sge, sge, num_sge,
&backend_dev->rdma_dev_res->stats.rx_bufs_len);
if (rc) {
complete_work(IBV_WC_GENERAL_ERR, rc, ctx);
@@ -652,13 +651,13 @@ void rdma_backend_post_recv(RdmaBackendDev *backend_dev,
}
atomic_inc(&backend_dev->rdma_dev_res->stats.missing_cqe);
- rdma_dev_res->stats.rx_bufs++;
+ backend_dev->rdma_dev_res->stats.rx_bufs++;
return;
err_dealloc_cqe_ctx:
backend_dev->rdma_dev_res->stats.rx_bufs_err++;
- rdma_rm_dealloc_cqe_ctx(rdma_dev_res, bctx_id);
+ rdma_rm_dealloc_cqe_ctx(backend_dev->rdma_dev_res, bctx_id);
err_free_bctx:
g_free(bctx);
diff --git a/hw/rdma/rdma_backend.h b/hw/rdma/rdma_backend.h
index cb5efa2a3a..5d507a1c41 100644
--- a/hw/rdma/rdma_backend.h
+++ b/hw/rdma/rdma_backend.h
@@ -111,7 +111,6 @@ void rdma_backend_post_send(RdmaBackendDev *backend_dev,
union ibv_gid *dgid, uint32_t dqpn, uint32_t dqkey,
void *ctx);
void rdma_backend_post_recv(RdmaBackendDev *backend_dev,
- RdmaDeviceResources *rdma_dev_res,
RdmaBackendQP *qp, uint8_t qp_type,
struct ibv_sge *sge, uint32_t num_sge, void *ctx);
diff --git a/hw/rdma/vmw/pvrdma_qp_ops.c b/hw/rdma/vmw/pvrdma_qp_ops.c
index 16db726dac..508d8fca3c 100644
--- a/hw/rdma/vmw/pvrdma_qp_ops.c
+++ b/hw/rdma/vmw/pvrdma_qp_ops.c
@@ -231,8 +231,7 @@ void pvrdma_qp_recv(PVRDMADev *dev, uint32_t qp_handle)
continue;
}
- rdma_backend_post_recv(&dev->backend_dev, &dev->rdma_dev_res,
- &qp->backend_qp, qp->qp_type,
+ rdma_backend_post_recv(&dev->backend_dev, &qp->backend_qp, qp->qp_type,
(struct ibv_sge *)&wqe->sge[0], wqe->hdr.num_sge,
comp_ctx);
--
2.17.2
^ permalink raw reply related [flat|nested] 17+ messages in thread
* [Qemu-devel] [PATCH v4 8/9] hw/pvrdma: Delete pvrdma_exit function
2019-03-03 20:33 [Qemu-devel] [PATCH v4 0/9] Misc fixes to pvrdma device Yuval Shaia
` (6 preceding siblings ...)
2019-03-03 20:33 ` [Qemu-devel] [PATCH v4 7/9] hw/pvrdma: Delete unneeded function argument Yuval Shaia
@ 2019-03-03 20:33 ` Yuval Shaia
2019-03-03 20:33 ` [Qemu-devel] [PATCH v4 9/9] hw/pvrdma: Unregister from shutdown notifier when device goes down Yuval Shaia
8 siblings, 0 replies; 17+ messages in thread
From: Yuval Shaia @ 2019-03-03 20:33 UTC (permalink / raw)
To: dgilbert, yuval.shaia, marcel.apfelbaum, armbru, qemu-devel
This hook is not called and was implemented by mistake.
Signed-off-by: Yuval Shaia <yuval.shaia@oracle.com>
Reviewed-by: Marcel Apfelbaum <marcel.apfelbaum@gmail.com>
---
hw/rdma/vmw/pvrdma_main.c | 6 ------
1 file changed, 6 deletions(-)
diff --git a/hw/rdma/vmw/pvrdma_main.c b/hw/rdma/vmw/pvrdma_main.c
index b795f80666..4e4a43eac4 100644
--- a/hw/rdma/vmw/pvrdma_main.c
+++ b/hw/rdma/vmw/pvrdma_main.c
@@ -641,11 +641,6 @@ out:
}
}
-static void pvrdma_exit(PCIDevice *pdev)
-{
- pvrdma_fini(pdev);
-}
-
static void pvrdma_class_init(ObjectClass *klass, void *data)
{
DeviceClass *dc = DEVICE_CLASS(klass);
@@ -653,7 +648,6 @@ static void pvrdma_class_init(ObjectClass *klass, void *data)
RdmaStatsProviderClass *ir = RDMA_STATS_PROVIDER_CLASS(klass);
k->realize = pvrdma_realize;
- k->exit = pvrdma_exit;
k->vendor_id = PCI_VENDOR_ID_VMWARE;
k->device_id = PCI_DEVICE_ID_VMWARE_PVRDMA;
k->revision = 0x00;
--
2.17.2
^ permalink raw reply related [flat|nested] 17+ messages in thread
* [Qemu-devel] [PATCH v4 9/9] hw/pvrdma: Unregister from shutdown notifier when device goes down
2019-03-03 20:33 [Qemu-devel] [PATCH v4 0/9] Misc fixes to pvrdma device Yuval Shaia
` (7 preceding siblings ...)
2019-03-03 20:33 ` [Qemu-devel] [PATCH v4 8/9] hw/pvrdma: Delete pvrdma_exit function Yuval Shaia
@ 2019-03-03 20:33 ` Yuval Shaia
8 siblings, 0 replies; 17+ messages in thread
From: Yuval Shaia @ 2019-03-03 20:33 UTC (permalink / raw)
To: dgilbert, yuval.shaia, marcel.apfelbaum, armbru, qemu-devel
This hook was installed to close the device when VM is going down.
After the device is closed there is no need to be informed on VM
shutdown.
Signed-off-by: Yuval Shaia <yuval.shaia@oracle.com>
Reviewed-by: Marcel Apfelbaum <marcel.apfelbaum@gmail.com>
---
hw/rdma/vmw/pvrdma_main.c | 2 ++
1 file changed, 2 insertions(+)
diff --git a/hw/rdma/vmw/pvrdma_main.c b/hw/rdma/vmw/pvrdma_main.c
index 4e4a43eac4..cb6eefe029 100644
--- a/hw/rdma/vmw/pvrdma_main.c
+++ b/hw/rdma/vmw/pvrdma_main.c
@@ -303,6 +303,8 @@ static void pvrdma_fini(PCIDevice *pdev)
{
PVRDMADev *dev = PVRDMA_DEV(pdev);
+ notifier_remove(&dev->shutdown_notifier);
+
pvrdma_qp_ops_fini();
rdma_backend_stop(&dev->backend_dev);
--
2.17.2
^ permalink raw reply related [flat|nested] 17+ messages in thread
* Re: [Qemu-devel] [PATCH v4 4/9] {hmp, hw/pvrdma}: Expose device internals via monitor interface
2019-03-03 20:33 ` [Qemu-devel] [PATCH v4 4/9] {hmp, hw/pvrdma}: Expose device internals via monitor interface Yuval Shaia
@ 2019-03-06 10:22 ` Yuval Shaia
2019-03-06 12:20 ` Dr. David Alan Gilbert
2019-03-07 9:50 ` Marcel Apfelbaum
2019-03-08 16:37 ` Markus Armbruster
2 siblings, 1 reply; 17+ messages in thread
From: Yuval Shaia @ 2019-03-06 10:22 UTC (permalink / raw)
To: dgilbert, marcel.apfelbaum, armbru, qemu-devel
On Sun, Mar 03, 2019 at 10:33:40PM +0200, Yuval Shaia wrote:
> Allow interrogating device internals through HMP interface.
> The exposed indicators can be used for troubleshooting by developers or
> sysadmin.
> There is no need to expose these attributes to a management system (e.x.
> libvirt) because (1) most of them are not "device-management' related
> info and (2) there is no guarantee the interface is stable.
>
> Signed-off-by: Yuval Shaia <yuval.shaia@oracle.com>
> ---
> hmp-commands-info.hx | 14 ++++++++
> hmp.c | 27 +++++++++++++++
> hmp.h | 1 +
> hw/rdma/Makefile.objs | 2 +-
> hw/rdma/rdma_backend.c | 70 +++++++++++++++++++++++++++++---------
> hw/rdma/rdma_hmp.c | 30 ++++++++++++++++
> hw/rdma/rdma_rm.c | 60 ++++++++++++++++++++++++++++++++
> hw/rdma/rdma_rm.h | 1 +
> hw/rdma/rdma_rm_defs.h | 27 ++++++++++++++-
> hw/rdma/vmw/pvrdma.h | 5 +++
> hw/rdma/vmw/pvrdma_main.c | 21 ++++++++++++
> include/hw/rdma/rdma_hmp.h | 40 ++++++++++++++++++++++
> 12 files changed, 279 insertions(+), 19 deletions(-)
> create mode 100644 hw/rdma/rdma_hmp.c
> create mode 100644 include/hw/rdma/rdma_hmp.h
>
> diff --git a/hmp-commands-info.hx b/hmp-commands-info.hx
> index cbee8b944d..c59444c461 100644
> --- a/hmp-commands-info.hx
> +++ b/hmp-commands-info.hx
> @@ -202,6 +202,20 @@ STEXI
> @item info pic
> @findex info pic
> Show PIC state.
> +ETEXI
> +
> + {
> + .name = "rdma",
> + .args_type = "",
> + .params = "",
> + .help = "show RDMA state",
> + .cmd = hmp_info_rdma,
> + },
> +
> +STEXI
> +@item info rdma
> +@findex info rdma
> +Show RDMA state.
> ETEXI
>
> {
> diff --git a/hmp.c b/hmp.c
> index 1e006eeb49..68511b2441 100644
> --- a/hmp.c
> +++ b/hmp.c
> @@ -51,6 +51,7 @@
> #include "qemu/error-report.h"
> #include "exec/ramlist.h"
> #include "hw/intc/intc.h"
> +#include "hw/rdma/rdma_hmp.h"
> #include "migration/snapshot.h"
> #include "migration/misc.h"
>
> @@ -968,6 +969,32 @@ void hmp_info_pic(Monitor *mon, const QDict *qdict)
> hmp_info_pic_foreach, mon);
> }
>
> +static int hmp_info_rdma_foreach(Object *obj, void *opaque)
> +{
> + RdmaStatsProvider *rdma;
> + RdmaStatsProviderClass *k;
> + Monitor *mon = opaque;
> +
> + if (object_dynamic_cast(obj, TYPE_RDMA_STATS_PROVIDER)) {
> + rdma = RDMA_STATS_PROVIDER(obj);
> + k = RDMA_STATS_PROVIDER_GET_CLASS(obj);
> + if (k->print_statistics) {
> + k->print_statistics(mon, rdma);
> + } else {
> + monitor_printf(mon, "RDMA statistics not available for %s.\n",
> + object_get_typename(obj));
> + }
> + }
> +
> + return 0;
> +}
> +
> +void hmp_info_rdma(Monitor *mon, const QDict *qdict)
> +{
> + object_child_foreach_recursive(object_get_root(),
> + hmp_info_rdma_foreach, mon);
> +}
> +
Hi Markus and David,
Is this implementation is acceptable by you?
Thanks,
Yuval
> void hmp_info_pci(Monitor *mon, const QDict *qdict)
> {
> PciInfoList *info_list, *info;
> diff --git a/hmp.h b/hmp.h
> index 5f1addcca2..666949afc3 100644
> --- a/hmp.h
> +++ b/hmp.h
> @@ -36,6 +36,7 @@ void hmp_info_spice(Monitor *mon, const QDict *qdict);
> void hmp_info_balloon(Monitor *mon, const QDict *qdict);
> void hmp_info_irq(Monitor *mon, const QDict *qdict);
> void hmp_info_pic(Monitor *mon, const QDict *qdict);
> +void hmp_info_rdma(Monitor *mon, const QDict *qdict);
> void hmp_info_pci(Monitor *mon, const QDict *qdict);
> void hmp_info_block_jobs(Monitor *mon, const QDict *qdict);
> void hmp_info_tpm(Monitor *mon, const QDict *qdict);
> diff --git a/hw/rdma/Makefile.objs b/hw/rdma/Makefile.objs
> index bd36cbf51c..dd59faff51 100644
> --- a/hw/rdma/Makefile.objs
> +++ b/hw/rdma/Makefile.objs
> @@ -1,5 +1,5 @@
> ifeq ($(CONFIG_PVRDMA),y)
> -obj-$(CONFIG_PCI) += rdma_utils.o rdma_backend.o rdma_rm.o
> +obj-$(CONFIG_PCI) += rdma_utils.o rdma_backend.o rdma_rm.o rdma_hmp.o
> obj-$(CONFIG_PCI) += vmw/pvrdma_dev_ring.o vmw/pvrdma_cmd.o \
> vmw/pvrdma_qp_ops.o vmw/pvrdma_main.o
> endif
> diff --git a/hw/rdma/rdma_backend.c b/hw/rdma/rdma_backend.c
> index 9679b842d1..bc2fefcf93 100644
> --- a/hw/rdma/rdma_backend.c
> +++ b/hw/rdma/rdma_backend.c
> @@ -64,9 +64,9 @@ static inline void complete_work(enum ibv_wc_status status, uint32_t vendor_err,
> comp_handler(ctx, &wc);
> }
>
> -static void rdma_poll_cq(RdmaDeviceResources *rdma_dev_res, struct ibv_cq *ibcq)
> +static int rdma_poll_cq(RdmaDeviceResources *rdma_dev_res, struct ibv_cq *ibcq)
> {
> - int i, ne;
> + int i, ne, total_ne = 0;
> BackendCtx *bctx;
> struct ibv_wc wc[2];
>
> @@ -89,12 +89,18 @@ static void rdma_poll_cq(RdmaDeviceResources *rdma_dev_res, struct ibv_cq *ibcq)
> rdma_rm_dealloc_cqe_ctx(rdma_dev_res, wc[i].wr_id);
> g_free(bctx);
> }
> + total_ne += ne;
> } while (ne > 0);
> + atomic_sub(&rdma_dev_res->stats.missing_cqe, total_ne);
> qemu_mutex_unlock(&rdma_dev_res->lock);
>
> if (ne < 0) {
> rdma_error_report("ibv_poll_cq fail, rc=%d, errno=%d", ne, errno);
> }
> +
> + rdma_dev_res->stats.completions += total_ne;
> +
> + return total_ne;
> }
>
> static void *comp_handler_thread(void *arg)
> @@ -122,6 +128,9 @@ static void *comp_handler_thread(void *arg)
> while (backend_dev->comp_thread.run) {
> do {
> rc = qemu_poll_ns(pfds, 1, THR_POLL_TO * (int64_t)SCALE_MS);
> + if (!rc) {
> + backend_dev->rdma_dev_res->stats.poll_cq_ppoll_to++;
> + }
> } while (!rc && backend_dev->comp_thread.run);
>
> if (backend_dev->comp_thread.run) {
> @@ -138,6 +147,7 @@ static void *comp_handler_thread(void *arg)
> errno);
> }
>
> + backend_dev->rdma_dev_res->stats.poll_cq_from_bk++;
> rdma_poll_cq(backend_dev->rdma_dev_res, ev_cq);
>
> ibv_ack_cq_events(ev_cq, 1);
> @@ -271,7 +281,13 @@ int rdma_backend_query_port(RdmaBackendDev *backend_dev,
>
> void rdma_backend_poll_cq(RdmaDeviceResources *rdma_dev_res, RdmaBackendCQ *cq)
> {
> - rdma_poll_cq(rdma_dev_res, cq->ibcq);
> + int polled;
> +
> + rdma_dev_res->stats.poll_cq_from_guest++;
> + polled = rdma_poll_cq(rdma_dev_res, cq->ibcq);
> + if (!polled) {
> + rdma_dev_res->stats.poll_cq_from_guest_empty++;
> + }
> }
>
> static GHashTable *ah_hash;
> @@ -333,7 +349,7 @@ static void ah_cache_init(void)
>
> static int build_host_sge_array(RdmaDeviceResources *rdma_dev_res,
> struct ibv_sge *dsge, struct ibv_sge *ssge,
> - uint8_t num_sge)
> + uint8_t num_sge, uint64_t *total_length)
> {
> RdmaRmMR *mr;
> int ssge_idx;
> @@ -349,6 +365,8 @@ static int build_host_sge_array(RdmaDeviceResources *rdma_dev_res,
> dsge->length = ssge[ssge_idx].length;
> dsge->lkey = rdma_backend_mr_lkey(&mr->backend_mr);
>
> + *total_length += dsge->length;
> +
> dsge++;
> }
>
> @@ -445,8 +463,10 @@ void rdma_backend_post_send(RdmaBackendDev *backend_dev,
> rc = mad_send(backend_dev, sgid_idx, sgid, sge, num_sge);
> if (rc) {
> complete_work(IBV_WC_GENERAL_ERR, VENDOR_ERR_MAD_SEND, ctx);
> + backend_dev->rdma_dev_res->stats.mad_tx_err++;
> } else {
> complete_work(IBV_WC_SUCCESS, 0, ctx);
> + backend_dev->rdma_dev_res->stats.mad_tx++;
> }
> }
> return;
> @@ -458,20 +478,21 @@ void rdma_backend_post_send(RdmaBackendDev *backend_dev,
> rc = rdma_rm_alloc_cqe_ctx(backend_dev->rdma_dev_res, &bctx_id, bctx);
> if (unlikely(rc)) {
> complete_work(IBV_WC_GENERAL_ERR, VENDOR_ERR_NOMEM, ctx);
> - goto out_free_bctx;
> + goto err_free_bctx;
> }
>
> - rc = build_host_sge_array(backend_dev->rdma_dev_res, new_sge, sge, num_sge);
> + rc = build_host_sge_array(backend_dev->rdma_dev_res, new_sge, sge, num_sge,
> + &backend_dev->rdma_dev_res->stats.tx_len);
> if (rc) {
> complete_work(IBV_WC_GENERAL_ERR, rc, ctx);
> - goto out_dealloc_cqe_ctx;
> + goto err_dealloc_cqe_ctx;
> }
>
> if (qp_type == IBV_QPT_UD) {
> wr.wr.ud.ah = create_ah(backend_dev, qp->ibpd, sgid_idx, dgid);
> if (!wr.wr.ud.ah) {
> complete_work(IBV_WC_GENERAL_ERR, VENDOR_ERR_FAIL_BACKEND, ctx);
> - goto out_dealloc_cqe_ctx;
> + goto err_dealloc_cqe_ctx;
> }
> wr.wr.ud.remote_qpn = dqpn;
> wr.wr.ud.remote_qkey = dqkey;
> @@ -488,15 +509,19 @@ void rdma_backend_post_send(RdmaBackendDev *backend_dev,
> rdma_error_report("ibv_post_send fail, qpn=0x%x, rc=%d, errno=%d",
> qp->ibqp->qp_num, rc, errno);
> complete_work(IBV_WC_GENERAL_ERR, VENDOR_ERR_FAIL_BACKEND, ctx);
> - goto out_dealloc_cqe_ctx;
> + goto err_dealloc_cqe_ctx;
> }
>
> + atomic_inc(&backend_dev->rdma_dev_res->stats.missing_cqe);
> + backend_dev->rdma_dev_res->stats.tx++;
> +
> return;
>
> -out_dealloc_cqe_ctx:
> +err_dealloc_cqe_ctx:
> + backend_dev->rdma_dev_res->stats.tx_err++;
> rdma_rm_dealloc_cqe_ctx(backend_dev->rdma_dev_res, bctx_id);
>
> -out_free_bctx:
> +err_free_bctx:
> g_free(bctx);
> }
>
> @@ -554,6 +579,9 @@ void rdma_backend_post_recv(RdmaBackendDev *backend_dev,
> rc = save_mad_recv_buffer(backend_dev, sge, num_sge, ctx);
> if (rc) {
> complete_work(IBV_WC_GENERAL_ERR, rc, ctx);
> + rdma_dev_res->stats.mad_rx_bufs_err++;
> + } else {
> + rdma_dev_res->stats.mad_rx_bufs++;
> }
> }
> return;
> @@ -565,13 +593,14 @@ void rdma_backend_post_recv(RdmaBackendDev *backend_dev,
> rc = rdma_rm_alloc_cqe_ctx(rdma_dev_res, &bctx_id, bctx);
> if (unlikely(rc)) {
> complete_work(IBV_WC_GENERAL_ERR, VENDOR_ERR_NOMEM, ctx);
> - goto out_free_bctx;
> + goto err_free_bctx;
> }
>
> - rc = build_host_sge_array(rdma_dev_res, new_sge, sge, num_sge);
> + rc = build_host_sge_array(rdma_dev_res, new_sge, sge, num_sge,
> + &backend_dev->rdma_dev_res->stats.rx_bufs_len);
> if (rc) {
> complete_work(IBV_WC_GENERAL_ERR, rc, ctx);
> - goto out_dealloc_cqe_ctx;
> + goto err_dealloc_cqe_ctx;
> }
>
> wr.num_sge = num_sge;
> @@ -582,15 +611,19 @@ void rdma_backend_post_recv(RdmaBackendDev *backend_dev,
> rdma_error_report("ibv_post_recv fail, qpn=0x%x, rc=%d, errno=%d",
> qp->ibqp->qp_num, rc, errno);
> complete_work(IBV_WC_GENERAL_ERR, VENDOR_ERR_FAIL_BACKEND, ctx);
> - goto out_dealloc_cqe_ctx;
> + goto err_dealloc_cqe_ctx;
> }
>
> + atomic_inc(&backend_dev->rdma_dev_res->stats.missing_cqe);
> + rdma_dev_res->stats.rx_bufs++;
> +
> return;
>
> -out_dealloc_cqe_ctx:
> +err_dealloc_cqe_ctx:
> + backend_dev->rdma_dev_res->stats.rx_bufs_err++;
> rdma_rm_dealloc_cqe_ctx(rdma_dev_res, bctx_id);
>
> -out_free_bctx:
> +err_free_bctx:
> g_free(bctx);
> }
>
> @@ -929,12 +962,14 @@ static void process_incoming_mad_req(RdmaBackendDev *backend_dev,
> bctx = rdma_rm_get_cqe_ctx(backend_dev->rdma_dev_res, cqe_ctx_id);
> if (unlikely(!bctx)) {
> rdma_error_report("No matching ctx for req %ld", cqe_ctx_id);
> + backend_dev->rdma_dev_res->stats.mad_rx_err++;
> return;
> }
>
> mad = rdma_pci_dma_map(backend_dev->dev, bctx->sge.addr,
> bctx->sge.length);
> if (!mad || bctx->sge.length < msg->umad_len + MAD_HDR_SIZE) {
> + backend_dev->rdma_dev_res->stats.mad_rx_err++;
> complete_work(IBV_WC_GENERAL_ERR, VENDOR_ERR_INV_MAD_BUFF,
> bctx->up_ctx);
> } else {
> @@ -949,6 +984,7 @@ static void process_incoming_mad_req(RdmaBackendDev *backend_dev,
> wc.byte_len = msg->umad_len;
> wc.status = IBV_WC_SUCCESS;
> wc.wc_flags = IBV_WC_GRH;
> + backend_dev->rdma_dev_res->stats.mad_rx++;
> comp_handler(bctx->up_ctx, &wc);
> }
>
> diff --git a/hw/rdma/rdma_hmp.c b/hw/rdma/rdma_hmp.c
> new file mode 100644
> index 0000000000..c5814473c5
> --- /dev/null
> +++ b/hw/rdma/rdma_hmp.c
> @@ -0,0 +1,30 @@
> +/*
> + * RDMA device: Human Monitor interface
> + *
> + * Copyright (C) 2018 Oracle
> + * Copyright (C) 2018 Red Hat Inc
> + *
> + * Authors:
> + * Yuval Shaia <yuval.shaia@oracle.com>
> + *
> + * This work is licensed under the terms of the GNU GPL, version 2 or later.
> + * See the COPYING file in the top-level directory.
> + *
> + */
> +
> +#include "qemu/osdep.h"
> +#include "hw/rdma/rdma_hmp.h"
> +#include "qemu/module.h"
> +
> +static const TypeInfo rdma_hmp_info = {
> + .name = TYPE_RDMA_STATS_PROVIDER,
> + .parent = TYPE_INTERFACE,
> + .class_size = sizeof(RdmaStatsProviderClass),
> +};
> +
> +static void rdma_hmp_register_types(void)
> +{
> + type_register_static(&rdma_hmp_info);
> +}
> +
> +type_init(rdma_hmp_register_types)
> diff --git a/hw/rdma/rdma_rm.c b/hw/rdma/rdma_rm.c
> index 14580ca379..e019de1a14 100644
> --- a/hw/rdma/rdma_rm.c
> +++ b/hw/rdma/rdma_rm.c
> @@ -16,6 +16,7 @@
> #include "qemu/osdep.h"
> #include "qapi/error.h"
> #include "cpu.h"
> +#include "monitor/monitor.h"
>
> #include "trace.h"
> #include "rdma_utils.h"
> @@ -26,6 +27,58 @@
> #define PG_DIR_SZ { TARGET_PAGE_SIZE / sizeof(__u64) }
> #define PG_TBL_SZ { TARGET_PAGE_SIZE / sizeof(__u64) }
>
> +void rdma_dump_device_counters(Monitor *mon, RdmaDeviceResources *dev_res)
> +{
> + monitor_printf(mon, "\ttx : %" PRId64 "\n",
> + dev_res->stats.tx);
> + monitor_printf(mon, "\ttx_len : %" PRId64 "\n",
> + dev_res->stats.tx_len);
> + monitor_printf(mon, "\ttx_err : %" PRId64 "\n",
> + dev_res->stats.tx_err);
> + monitor_printf(mon, "\trx_bufs : %" PRId64 "\n",
> + dev_res->stats.rx_bufs);
> + monitor_printf(mon, "\trx_bufs_len : %" PRId64 "\n",
> + dev_res->stats.rx_bufs_len);
> + monitor_printf(mon, "\trx_bufs_err : %" PRId64 "\n",
> + dev_res->stats.rx_bufs_err);
> + monitor_printf(mon, "\tcomps : %" PRId64 "\n",
> + dev_res->stats.completions);
> + monitor_printf(mon, "\tmissing_comps : %" PRId32 "\n",
> + dev_res->stats.missing_cqe);
> + monitor_printf(mon, "\tpoll_cq (bk) : %" PRId64 "\n",
> + dev_res->stats.poll_cq_from_bk);
> + monitor_printf(mon, "\tpoll_cq_ppoll_to : %" PRId64 "\n",
> + dev_res->stats.poll_cq_ppoll_to);
> + monitor_printf(mon, "\tpoll_cq (fe) : %" PRId64 "\n",
> + dev_res->stats.poll_cq_from_guest);
> + monitor_printf(mon, "\tpoll_cq_empty : %" PRId64 "\n",
> + dev_res->stats.poll_cq_from_guest_empty);
> + monitor_printf(mon, "\tmad_tx : %" PRId64 "\n",
> + dev_res->stats.mad_tx);
> + monitor_printf(mon, "\tmad_tx_err : %" PRId64 "\n",
> + dev_res->stats.mad_tx_err);
> + monitor_printf(mon, "\tmad_rx : %" PRId64 "\n",
> + dev_res->stats.mad_rx);
> + monitor_printf(mon, "\tmad_rx_err : %" PRId64 "\n",
> + dev_res->stats.mad_rx_err);
> + monitor_printf(mon, "\tmad_rx_bufs : %" PRId64 "\n",
> + dev_res->stats.mad_rx_bufs);
> + monitor_printf(mon, "\tmad_rx_bufs_err : %" PRId64 "\n",
> + dev_res->stats.mad_rx_bufs_err);
> + monitor_printf(mon, "\tPDs : %" PRId32 "\n",
> + dev_res->pd_tbl.used);
> + monitor_printf(mon, "\tMRs : %" PRId32 "\n",
> + dev_res->mr_tbl.used);
> + monitor_printf(mon, "\tUCs : %" PRId32 "\n",
> + dev_res->uc_tbl.used);
> + monitor_printf(mon, "\tQPs : %" PRId32 "\n",
> + dev_res->qp_tbl.used);
> + monitor_printf(mon, "\tCQs : %" PRId32 "\n",
> + dev_res->cq_tbl.used);
> + monitor_printf(mon, "\tCEQ_CTXs : %" PRId32 "\n",
> + dev_res->cqe_ctx_tbl.used);
> +}
> +
> static inline void res_tbl_init(const char *name, RdmaRmResTbl *tbl,
> uint32_t tbl_sz, uint32_t res_sz)
> {
> @@ -37,6 +90,7 @@ static inline void res_tbl_init(const char *name, RdmaRmResTbl *tbl,
> tbl->bitmap = bitmap_new(tbl_sz);
> tbl->tbl_sz = tbl_sz;
> tbl->res_sz = res_sz;
> + tbl->used = 0;
> qemu_mutex_init(&tbl->lock);
> }
>
> @@ -76,6 +130,8 @@ static inline void *rdma_res_tbl_alloc(RdmaRmResTbl *tbl, uint32_t *handle)
>
> set_bit(*handle, tbl->bitmap);
>
> + tbl->used++;
> +
> qemu_mutex_unlock(&tbl->lock);
>
> memset(tbl->tbl + *handle * tbl->res_sz, 0, tbl->res_sz);
> @@ -93,6 +149,7 @@ static inline void rdma_res_tbl_dealloc(RdmaRmResTbl *tbl, uint32_t handle)
>
> if (handle < tbl->tbl_sz) {
> clear_bit(handle, tbl->bitmap);
> + tbl->used--;
> }
>
> qemu_mutex_unlock(&tbl->lock);
> @@ -620,6 +677,9 @@ int rdma_rm_init(RdmaDeviceResources *dev_res, struct ibv_device_attr *dev_attr,
>
> qemu_mutex_init(&dev_res->lock);
>
> + memset(&dev_res->stats, 0, sizeof(dev_res->stats));
> + atomic_set(&dev_res->stats.missing_cqe, 0);
> +
> return 0;
> }
>
> diff --git a/hw/rdma/rdma_rm.h b/hw/rdma/rdma_rm.h
> index 9ec87d2667..a527d306f6 100644
> --- a/hw/rdma/rdma_rm.h
> +++ b/hw/rdma/rdma_rm.h
> @@ -20,6 +20,7 @@
> #include "rdma_backend_defs.h"
> #include "rdma_rm_defs.h"
>
> +void rdma_dump_device_counters(Monitor *mon, RdmaDeviceResources *dev_res);
> int rdma_rm_init(RdmaDeviceResources *dev_res, struct ibv_device_attr *dev_attr,
> Error **errp);
> void rdma_rm_fini(RdmaDeviceResources *dev_res, RdmaBackendDev *backend_dev,
> diff --git a/hw/rdma/rdma_rm_defs.h b/hw/rdma/rdma_rm_defs.h
> index f0ee1f3072..4b8d704cfe 100644
> --- a/hw/rdma/rdma_rm_defs.h
> +++ b/hw/rdma/rdma_rm_defs.h
> @@ -34,7 +34,9 @@
> #define MAX_QP_INIT_RD_ATOM 16
> #define MAX_AH 64
>
> -#define MAX_RM_TBL_NAME 16
> +#define MAX_RM_TBL_NAME 16
> +#define MAX_CONSEQ_EMPTY_POLL_CQ 4096 /* considered as error above this */
> +
> typedef struct RdmaRmResTbl {
> char name[MAX_RM_TBL_NAME];
> QemuMutex lock;
> @@ -42,6 +44,7 @@ typedef struct RdmaRmResTbl {
> size_t tbl_sz;
> size_t res_sz;
> void *tbl;
> + uint32_t used; /* number of used entries in the table */
> } RdmaRmResTbl;
>
> typedef struct RdmaRmPD {
> @@ -96,6 +99,27 @@ typedef struct RdmaRmPort {
> enum ibv_port_state state;
> } RdmaRmPort;
>
> +typedef struct RdmaRmStats {
> + uint64_t tx;
> + uint64_t tx_len;
> + uint64_t tx_err;
> + uint64_t rx_bufs;
> + uint64_t rx_bufs_len;
> + uint64_t rx_bufs_err;
> + uint64_t completions;
> + uint64_t mad_tx;
> + uint64_t mad_tx_err;
> + uint64_t mad_rx;
> + uint64_t mad_rx_err;
> + uint64_t mad_rx_bufs;
> + uint64_t mad_rx_bufs_err;
> + uint64_t poll_cq_from_bk;
> + uint64_t poll_cq_from_guest;
> + uint64_t poll_cq_from_guest_empty;
> + uint64_t poll_cq_ppoll_to;
> + uint32_t missing_cqe;
> +} RdmaRmStats;
> +
> typedef struct RdmaDeviceResources {
> RdmaRmPort port;
> RdmaRmResTbl pd_tbl;
> @@ -106,6 +130,7 @@ typedef struct RdmaDeviceResources {
> RdmaRmResTbl cqe_ctx_tbl;
> GHashTable *qp_hash; /* Keeps mapping between real and emulated */
> QemuMutex lock;
> + RdmaRmStats stats;
> } RdmaDeviceResources;
>
> #endif
> diff --git a/hw/rdma/vmw/pvrdma.h b/hw/rdma/vmw/pvrdma.h
> index 0879224957..167706ec2c 100644
> --- a/hw/rdma/vmw/pvrdma.h
> +++ b/hw/rdma/vmw/pvrdma.h
> @@ -70,6 +70,10 @@ typedef struct DSRInfo {
> PvrdmaRing cq;
> } DSRInfo;
>
> +typedef struct PVRDMADevStats {
> + uint64_t commands;
> +} PVRDMADevStats;
> +
> typedef struct PVRDMADev {
> PCIDevice parent_obj;
> MemoryRegion msix;
> @@ -89,6 +93,7 @@ typedef struct PVRDMADev {
> CharBackend mad_chr;
> VMXNET3State *func0;
> Notifier shutdown_notifier;
> + PVRDMADevStats stats;
> } PVRDMADev;
> #define PVRDMA_DEV(dev) OBJECT_CHECK(PVRDMADev, (dev), PVRDMA_HW_NAME)
>
> diff --git a/hw/rdma/vmw/pvrdma_main.c b/hw/rdma/vmw/pvrdma_main.c
> index b6061f4b6e..659331ac93 100644
> --- a/hw/rdma/vmw/pvrdma_main.c
> +++ b/hw/rdma/vmw/pvrdma_main.c
> @@ -25,6 +25,8 @@
> #include "cpu.h"
> #include "trace.h"
> #include "sysemu/sysemu.h"
> +#include "monitor/monitor.h"
> +#include "hw/rdma/rdma_hmp.h"
>
> #include "../rdma_rm.h"
> #include "../rdma_backend.h"
> @@ -55,6 +57,18 @@ static Property pvrdma_dev_properties[] = {
> DEFINE_PROP_END_OF_LIST(),
> };
>
> +static void pvrdma_print_statistics(Monitor *mon, RdmaStatsProvider *obj)
> +{
> + PVRDMADev *dev = PVRDMA_DEV(obj);
> + PCIDevice *pdev = PCI_DEVICE(dev);
> +
> + monitor_printf(mon, "%s, %x.%x\n", pdev->name, PCI_SLOT(pdev->devfn),
> + PCI_FUNC(pdev->devfn));
> + monitor_printf(mon, "\tcommands : %" PRId64 "\n",
> + dev->stats.commands);
> + rdma_dump_device_counters(mon, &dev->rdma_dev_res);
> +}
> +
> static void free_dev_ring(PCIDevice *pci_dev, PvrdmaRing *ring,
> void *ring_state)
> {
> @@ -394,6 +408,7 @@ static void pvrdma_regs_write(void *opaque, hwaddr addr, uint64_t val,
> if (val == 0) {
> trace_pvrdma_regs_write(addr, val, "REQUEST", "");
> pvrdma_exec_cmd(dev);
> + dev->stats.commands++;
> }
> break;
> default:
> @@ -612,6 +627,8 @@ static void pvrdma_realize(PCIDevice *pdev, Error **errp)
> goto out;
> }
>
> + memset(&dev->stats, 0, sizeof(dev->stats));
> +
> dev->shutdown_notifier.notify = pvrdma_shutdown_notifier;
> qemu_register_shutdown_notifier(&dev->shutdown_notifier);
>
> @@ -631,6 +648,7 @@ static void pvrdma_class_init(ObjectClass *klass, void *data)
> {
> DeviceClass *dc = DEVICE_CLASS(klass);
> PCIDeviceClass *k = PCI_DEVICE_CLASS(klass);
> + RdmaStatsProviderClass *ir = RDMA_STATS_PROVIDER_CLASS(klass);
>
> k->realize = pvrdma_realize;
> k->exit = pvrdma_exit;
> @@ -642,6 +660,8 @@ static void pvrdma_class_init(ObjectClass *klass, void *data)
> dc->desc = "RDMA Device";
> dc->props = pvrdma_dev_properties;
> set_bit(DEVICE_CATEGORY_NETWORK, dc->categories);
> +
> + ir->print_statistics = pvrdma_print_statistics;
> }
>
> static const TypeInfo pvrdma_info = {
> @@ -651,6 +671,7 @@ static const TypeInfo pvrdma_info = {
> .class_init = pvrdma_class_init,
> .interfaces = (InterfaceInfo[]) {
> { INTERFACE_CONVENTIONAL_PCI_DEVICE },
> + { TYPE_RDMA_STATS_PROVIDER },
> { }
> }
> };
> diff --git a/include/hw/rdma/rdma_hmp.h b/include/hw/rdma/rdma_hmp.h
> new file mode 100644
> index 0000000000..dd23f2bc84
> --- /dev/null
> +++ b/include/hw/rdma/rdma_hmp.h
> @@ -0,0 +1,40 @@
> +/*
> + * RDMA device: Human Monitor interface
> + *
> + * Copyright (C) 2019 Oracle
> + * Copyright (C) 2019 Red Hat Inc
> + *
> + * Authors:
> + * Yuval Shaia <yuval.shaia@oracle.com>
> + *
> + * This work is licensed under the terms of the GNU GPL, version 2 or later.
> + * See the COPYING file in the top-level directory.
> + *
> + */
> +
> +#ifndef RDMA_HMP_H
> +#define RDMA_HMP_H
> +
> +#include "qom/object.h"
> +
> +#define TYPE_RDMA_STATS_PROVIDER "rdma"
> +
> +#define RDMA_STATS_PROVIDER_CLASS(klass) \
> + OBJECT_CLASS_CHECK(RdmaStatsProviderClass, (klass), \
> + TYPE_RDMA_STATS_PROVIDER)
> +#define RDMA_STATS_PROVIDER_GET_CLASS(obj) \
> + OBJECT_GET_CLASS(RdmaStatsProviderClass, (obj), \
> + TYPE_RDMA_STATS_PROVIDER)
> +#define RDMA_STATS_PROVIDER(obj) \
> + INTERFACE_CHECK(RdmaStatsProvider, (obj), \
> + TYPE_RDMA_STATS_PROVIDER)
> +
> +typedef struct RdmaStatsProvider RdmaStatsProvider;
> +
> +typedef struct RdmaStatsProviderClass {
> + InterfaceClass parent;
> +
> + void (*print_statistics)(Monitor *mon, RdmaStatsProvider *obj);
> +} RdmaStatsProviderClass;
> +
> +#endif
> --
> 2.17.2
>
>
^ permalink raw reply [flat|nested] 17+ messages in thread
* Re: [Qemu-devel] [PATCH v4 4/9] {hmp, hw/pvrdma}: Expose device internals via monitor interface
2019-03-06 10:22 ` Yuval Shaia
@ 2019-03-06 12:20 ` Dr. David Alan Gilbert
2019-03-06 12:46 ` Yuval Shaia
0 siblings, 1 reply; 17+ messages in thread
From: Dr. David Alan Gilbert @ 2019-03-06 12:20 UTC (permalink / raw)
To: Yuval Shaia; +Cc: marcel.apfelbaum, armbru, qemu-devel
* Yuval Shaia (yuval.shaia@oracle.com) wrote:
> On Sun, Mar 03, 2019 at 10:33:40PM +0200, Yuval Shaia wrote:
> > Allow interrogating device internals through HMP interface.
> > The exposed indicators can be used for troubleshooting by developers or
> > sysadmin.
> > There is no need to expose these attributes to a management system (e.x.
> > libvirt) because (1) most of them are not "device-management' related
> > info and (2) there is no guarantee the interface is stable.
> >
> > Signed-off-by: Yuval Shaia <yuval.shaia@oracle.com>
> > ---
> > hmp-commands-info.hx | 14 ++++++++
> > hmp.c | 27 +++++++++++++++
> > hmp.h | 1 +
> > hw/rdma/Makefile.objs | 2 +-
> > hw/rdma/rdma_backend.c | 70 +++++++++++++++++++++++++++++---------
> > hw/rdma/rdma_hmp.c | 30 ++++++++++++++++
> > hw/rdma/rdma_rm.c | 60 ++++++++++++++++++++++++++++++++
> > hw/rdma/rdma_rm.h | 1 +
> > hw/rdma/rdma_rm_defs.h | 27 ++++++++++++++-
> > hw/rdma/vmw/pvrdma.h | 5 +++
> > hw/rdma/vmw/pvrdma_main.c | 21 ++++++++++++
> > include/hw/rdma/rdma_hmp.h | 40 ++++++++++++++++++++++
> > 12 files changed, 279 insertions(+), 19 deletions(-)
> > create mode 100644 hw/rdma/rdma_hmp.c
> > create mode 100644 include/hw/rdma/rdma_hmp.h
> >
> > diff --git a/hmp-commands-info.hx b/hmp-commands-info.hx
> > index cbee8b944d..c59444c461 100644
> > --- a/hmp-commands-info.hx
> > +++ b/hmp-commands-info.hx
> > @@ -202,6 +202,20 @@ STEXI
> > @item info pic
> > @findex info pic
> > Show PIC state.
> > +ETEXI
> > +
> > + {
> > + .name = "rdma",
> > + .args_type = "",
> > + .params = "",
> > + .help = "show RDMA state",
> > + .cmd = hmp_info_rdma,
> > + },
> > +
> > +STEXI
> > +@item info rdma
> > +@findex info rdma
> > +Show RDMA state.
> > ETEXI
> >
> > {
> > diff --git a/hmp.c b/hmp.c
> > index 1e006eeb49..68511b2441 100644
> > --- a/hmp.c
> > +++ b/hmp.c
> > @@ -51,6 +51,7 @@
> > #include "qemu/error-report.h"
> > #include "exec/ramlist.h"
> > #include "hw/intc/intc.h"
> > +#include "hw/rdma/rdma_hmp.h"
> > #include "migration/snapshot.h"
> > #include "migration/misc.h"
> >
> > @@ -968,6 +969,32 @@ void hmp_info_pic(Monitor *mon, const QDict *qdict)
> > hmp_info_pic_foreach, mon);
> > }
> >
> > +static int hmp_info_rdma_foreach(Object *obj, void *opaque)
> > +{
> > + RdmaStatsProvider *rdma;
> > + RdmaStatsProviderClass *k;
> > + Monitor *mon = opaque;
> > +
> > + if (object_dynamic_cast(obj, TYPE_RDMA_STATS_PROVIDER)) {
> > + rdma = RDMA_STATS_PROVIDER(obj);
> > + k = RDMA_STATS_PROVIDER_GET_CLASS(obj);
> > + if (k->print_statistics) {
> > + k->print_statistics(mon, rdma);
> > + } else {
> > + monitor_printf(mon, "RDMA statistics not available for %s.\n",
> > + object_get_typename(obj));
> > + }
> > + }
> > +
> > + return 0;
> > +}
> > +
> > +void hmp_info_rdma(Monitor *mon, const QDict *qdict)
> > +{
> > + object_child_foreach_recursive(object_get_root(),
> > + hmp_info_rdma_foreach, mon);
> > +}
> > +
>
> Hi Markus and David,
> Is this implementation is acceptable by you?
I think from the HMP side I'm fine, so:
Acked-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
I would say that there's really two parts to this patch, all the places
you change your code to gather the stats, and the separate pieces that
wire it into HMP; if you had to repost for another reason I'd
split the patch like that.
Dave
> Thanks,
> Yuval
>
>
> > void hmp_info_pci(Monitor *mon, const QDict *qdict)
> > {
> > PciInfoList *info_list, *info;
> > diff --git a/hmp.h b/hmp.h
> > index 5f1addcca2..666949afc3 100644
> > --- a/hmp.h
> > +++ b/hmp.h
> > @@ -36,6 +36,7 @@ void hmp_info_spice(Monitor *mon, const QDict *qdict);
> > void hmp_info_balloon(Monitor *mon, const QDict *qdict);
> > void hmp_info_irq(Monitor *mon, const QDict *qdict);
> > void hmp_info_pic(Monitor *mon, const QDict *qdict);
> > +void hmp_info_rdma(Monitor *mon, const QDict *qdict);
> > void hmp_info_pci(Monitor *mon, const QDict *qdict);
> > void hmp_info_block_jobs(Monitor *mon, const QDict *qdict);
> > void hmp_info_tpm(Monitor *mon, const QDict *qdict);
> > diff --git a/hw/rdma/Makefile.objs b/hw/rdma/Makefile.objs
> > index bd36cbf51c..dd59faff51 100644
> > --- a/hw/rdma/Makefile.objs
> > +++ b/hw/rdma/Makefile.objs
> > @@ -1,5 +1,5 @@
> > ifeq ($(CONFIG_PVRDMA),y)
> > -obj-$(CONFIG_PCI) += rdma_utils.o rdma_backend.o rdma_rm.o
> > +obj-$(CONFIG_PCI) += rdma_utils.o rdma_backend.o rdma_rm.o rdma_hmp.o
> > obj-$(CONFIG_PCI) += vmw/pvrdma_dev_ring.o vmw/pvrdma_cmd.o \
> > vmw/pvrdma_qp_ops.o vmw/pvrdma_main.o
> > endif
> > diff --git a/hw/rdma/rdma_backend.c b/hw/rdma/rdma_backend.c
> > index 9679b842d1..bc2fefcf93 100644
> > --- a/hw/rdma/rdma_backend.c
> > +++ b/hw/rdma/rdma_backend.c
> > @@ -64,9 +64,9 @@ static inline void complete_work(enum ibv_wc_status status, uint32_t vendor_err,
> > comp_handler(ctx, &wc);
> > }
> >
> > -static void rdma_poll_cq(RdmaDeviceResources *rdma_dev_res, struct ibv_cq *ibcq)
> > +static int rdma_poll_cq(RdmaDeviceResources *rdma_dev_res, struct ibv_cq *ibcq)
> > {
> > - int i, ne;
> > + int i, ne, total_ne = 0;
> > BackendCtx *bctx;
> > struct ibv_wc wc[2];
> >
> > @@ -89,12 +89,18 @@ static void rdma_poll_cq(RdmaDeviceResources *rdma_dev_res, struct ibv_cq *ibcq)
> > rdma_rm_dealloc_cqe_ctx(rdma_dev_res, wc[i].wr_id);
> > g_free(bctx);
> > }
> > + total_ne += ne;
> > } while (ne > 0);
> > + atomic_sub(&rdma_dev_res->stats.missing_cqe, total_ne);
> > qemu_mutex_unlock(&rdma_dev_res->lock);
> >
> > if (ne < 0) {
> > rdma_error_report("ibv_poll_cq fail, rc=%d, errno=%d", ne, errno);
> > }
> > +
> > + rdma_dev_res->stats.completions += total_ne;
> > +
> > + return total_ne;
> > }
> >
> > static void *comp_handler_thread(void *arg)
> > @@ -122,6 +128,9 @@ static void *comp_handler_thread(void *arg)
> > while (backend_dev->comp_thread.run) {
> > do {
> > rc = qemu_poll_ns(pfds, 1, THR_POLL_TO * (int64_t)SCALE_MS);
> > + if (!rc) {
> > + backend_dev->rdma_dev_res->stats.poll_cq_ppoll_to++;
> > + }
> > } while (!rc && backend_dev->comp_thread.run);
> >
> > if (backend_dev->comp_thread.run) {
> > @@ -138,6 +147,7 @@ static void *comp_handler_thread(void *arg)
> > errno);
> > }
> >
> > + backend_dev->rdma_dev_res->stats.poll_cq_from_bk++;
> > rdma_poll_cq(backend_dev->rdma_dev_res, ev_cq);
> >
> > ibv_ack_cq_events(ev_cq, 1);
> > @@ -271,7 +281,13 @@ int rdma_backend_query_port(RdmaBackendDev *backend_dev,
> >
> > void rdma_backend_poll_cq(RdmaDeviceResources *rdma_dev_res, RdmaBackendCQ *cq)
> > {
> > - rdma_poll_cq(rdma_dev_res, cq->ibcq);
> > + int polled;
> > +
> > + rdma_dev_res->stats.poll_cq_from_guest++;
> > + polled = rdma_poll_cq(rdma_dev_res, cq->ibcq);
> > + if (!polled) {
> > + rdma_dev_res->stats.poll_cq_from_guest_empty++;
> > + }
> > }
> >
> > static GHashTable *ah_hash;
> > @@ -333,7 +349,7 @@ static void ah_cache_init(void)
> >
> > static int build_host_sge_array(RdmaDeviceResources *rdma_dev_res,
> > struct ibv_sge *dsge, struct ibv_sge *ssge,
> > - uint8_t num_sge)
> > + uint8_t num_sge, uint64_t *total_length)
> > {
> > RdmaRmMR *mr;
> > int ssge_idx;
> > @@ -349,6 +365,8 @@ static int build_host_sge_array(RdmaDeviceResources *rdma_dev_res,
> > dsge->length = ssge[ssge_idx].length;
> > dsge->lkey = rdma_backend_mr_lkey(&mr->backend_mr);
> >
> > + *total_length += dsge->length;
> > +
> > dsge++;
> > }
> >
> > @@ -445,8 +463,10 @@ void rdma_backend_post_send(RdmaBackendDev *backend_dev,
> > rc = mad_send(backend_dev, sgid_idx, sgid, sge, num_sge);
> > if (rc) {
> > complete_work(IBV_WC_GENERAL_ERR, VENDOR_ERR_MAD_SEND, ctx);
> > + backend_dev->rdma_dev_res->stats.mad_tx_err++;
> > } else {
> > complete_work(IBV_WC_SUCCESS, 0, ctx);
> > + backend_dev->rdma_dev_res->stats.mad_tx++;
> > }
> > }
> > return;
> > @@ -458,20 +478,21 @@ void rdma_backend_post_send(RdmaBackendDev *backend_dev,
> > rc = rdma_rm_alloc_cqe_ctx(backend_dev->rdma_dev_res, &bctx_id, bctx);
> > if (unlikely(rc)) {
> > complete_work(IBV_WC_GENERAL_ERR, VENDOR_ERR_NOMEM, ctx);
> > - goto out_free_bctx;
> > + goto err_free_bctx;
> > }
> >
> > - rc = build_host_sge_array(backend_dev->rdma_dev_res, new_sge, sge, num_sge);
> > + rc = build_host_sge_array(backend_dev->rdma_dev_res, new_sge, sge, num_sge,
> > + &backend_dev->rdma_dev_res->stats.tx_len);
> > if (rc) {
> > complete_work(IBV_WC_GENERAL_ERR, rc, ctx);
> > - goto out_dealloc_cqe_ctx;
> > + goto err_dealloc_cqe_ctx;
> > }
> >
> > if (qp_type == IBV_QPT_UD) {
> > wr.wr.ud.ah = create_ah(backend_dev, qp->ibpd, sgid_idx, dgid);
> > if (!wr.wr.ud.ah) {
> > complete_work(IBV_WC_GENERAL_ERR, VENDOR_ERR_FAIL_BACKEND, ctx);
> > - goto out_dealloc_cqe_ctx;
> > + goto err_dealloc_cqe_ctx;
> > }
> > wr.wr.ud.remote_qpn = dqpn;
> > wr.wr.ud.remote_qkey = dqkey;
> > @@ -488,15 +509,19 @@ void rdma_backend_post_send(RdmaBackendDev *backend_dev,
> > rdma_error_report("ibv_post_send fail, qpn=0x%x, rc=%d, errno=%d",
> > qp->ibqp->qp_num, rc, errno);
> > complete_work(IBV_WC_GENERAL_ERR, VENDOR_ERR_FAIL_BACKEND, ctx);
> > - goto out_dealloc_cqe_ctx;
> > + goto err_dealloc_cqe_ctx;
> > }
> >
> > + atomic_inc(&backend_dev->rdma_dev_res->stats.missing_cqe);
> > + backend_dev->rdma_dev_res->stats.tx++;
> > +
> > return;
> >
> > -out_dealloc_cqe_ctx:
> > +err_dealloc_cqe_ctx:
> > + backend_dev->rdma_dev_res->stats.tx_err++;
> > rdma_rm_dealloc_cqe_ctx(backend_dev->rdma_dev_res, bctx_id);
> >
> > -out_free_bctx:
> > +err_free_bctx:
> > g_free(bctx);
> > }
> >
> > @@ -554,6 +579,9 @@ void rdma_backend_post_recv(RdmaBackendDev *backend_dev,
> > rc = save_mad_recv_buffer(backend_dev, sge, num_sge, ctx);
> > if (rc) {
> > complete_work(IBV_WC_GENERAL_ERR, rc, ctx);
> > + rdma_dev_res->stats.mad_rx_bufs_err++;
> > + } else {
> > + rdma_dev_res->stats.mad_rx_bufs++;
> > }
> > }
> > return;
> > @@ -565,13 +593,14 @@ void rdma_backend_post_recv(RdmaBackendDev *backend_dev,
> > rc = rdma_rm_alloc_cqe_ctx(rdma_dev_res, &bctx_id, bctx);
> > if (unlikely(rc)) {
> > complete_work(IBV_WC_GENERAL_ERR, VENDOR_ERR_NOMEM, ctx);
> > - goto out_free_bctx;
> > + goto err_free_bctx;
> > }
> >
> > - rc = build_host_sge_array(rdma_dev_res, new_sge, sge, num_sge);
> > + rc = build_host_sge_array(rdma_dev_res, new_sge, sge, num_sge,
> > + &backend_dev->rdma_dev_res->stats.rx_bufs_len);
> > if (rc) {
> > complete_work(IBV_WC_GENERAL_ERR, rc, ctx);
> > - goto out_dealloc_cqe_ctx;
> > + goto err_dealloc_cqe_ctx;
> > }
> >
> > wr.num_sge = num_sge;
> > @@ -582,15 +611,19 @@ void rdma_backend_post_recv(RdmaBackendDev *backend_dev,
> > rdma_error_report("ibv_post_recv fail, qpn=0x%x, rc=%d, errno=%d",
> > qp->ibqp->qp_num, rc, errno);
> > complete_work(IBV_WC_GENERAL_ERR, VENDOR_ERR_FAIL_BACKEND, ctx);
> > - goto out_dealloc_cqe_ctx;
> > + goto err_dealloc_cqe_ctx;
> > }
> >
> > + atomic_inc(&backend_dev->rdma_dev_res->stats.missing_cqe);
> > + rdma_dev_res->stats.rx_bufs++;
> > +
> > return;
> >
> > -out_dealloc_cqe_ctx:
> > +err_dealloc_cqe_ctx:
> > + backend_dev->rdma_dev_res->stats.rx_bufs_err++;
> > rdma_rm_dealloc_cqe_ctx(rdma_dev_res, bctx_id);
> >
> > -out_free_bctx:
> > +err_free_bctx:
> > g_free(bctx);
> > }
> >
> > @@ -929,12 +962,14 @@ static void process_incoming_mad_req(RdmaBackendDev *backend_dev,
> > bctx = rdma_rm_get_cqe_ctx(backend_dev->rdma_dev_res, cqe_ctx_id);
> > if (unlikely(!bctx)) {
> > rdma_error_report("No matching ctx for req %ld", cqe_ctx_id);
> > + backend_dev->rdma_dev_res->stats.mad_rx_err++;
> > return;
> > }
> >
> > mad = rdma_pci_dma_map(backend_dev->dev, bctx->sge.addr,
> > bctx->sge.length);
> > if (!mad || bctx->sge.length < msg->umad_len + MAD_HDR_SIZE) {
> > + backend_dev->rdma_dev_res->stats.mad_rx_err++;
> > complete_work(IBV_WC_GENERAL_ERR, VENDOR_ERR_INV_MAD_BUFF,
> > bctx->up_ctx);
> > } else {
> > @@ -949,6 +984,7 @@ static void process_incoming_mad_req(RdmaBackendDev *backend_dev,
> > wc.byte_len = msg->umad_len;
> > wc.status = IBV_WC_SUCCESS;
> > wc.wc_flags = IBV_WC_GRH;
> > + backend_dev->rdma_dev_res->stats.mad_rx++;
> > comp_handler(bctx->up_ctx, &wc);
> > }
> >
> > diff --git a/hw/rdma/rdma_hmp.c b/hw/rdma/rdma_hmp.c
> > new file mode 100644
> > index 0000000000..c5814473c5
> > --- /dev/null
> > +++ b/hw/rdma/rdma_hmp.c
> > @@ -0,0 +1,30 @@
> > +/*
> > + * RDMA device: Human Monitor interface
> > + *
> > + * Copyright (C) 2018 Oracle
> > + * Copyright (C) 2018 Red Hat Inc
> > + *
> > + * Authors:
> > + * Yuval Shaia <yuval.shaia@oracle.com>
> > + *
> > + * This work is licensed under the terms of the GNU GPL, version 2 or later.
> > + * See the COPYING file in the top-level directory.
> > + *
> > + */
> > +
> > +#include "qemu/osdep.h"
> > +#include "hw/rdma/rdma_hmp.h"
> > +#include "qemu/module.h"
> > +
> > +static const TypeInfo rdma_hmp_info = {
> > + .name = TYPE_RDMA_STATS_PROVIDER,
> > + .parent = TYPE_INTERFACE,
> > + .class_size = sizeof(RdmaStatsProviderClass),
> > +};
> > +
> > +static void rdma_hmp_register_types(void)
> > +{
> > + type_register_static(&rdma_hmp_info);
> > +}
> > +
> > +type_init(rdma_hmp_register_types)
> > diff --git a/hw/rdma/rdma_rm.c b/hw/rdma/rdma_rm.c
> > index 14580ca379..e019de1a14 100644
> > --- a/hw/rdma/rdma_rm.c
> > +++ b/hw/rdma/rdma_rm.c
> > @@ -16,6 +16,7 @@
> > #include "qemu/osdep.h"
> > #include "qapi/error.h"
> > #include "cpu.h"
> > +#include "monitor/monitor.h"
> >
> > #include "trace.h"
> > #include "rdma_utils.h"
> > @@ -26,6 +27,58 @@
> > #define PG_DIR_SZ { TARGET_PAGE_SIZE / sizeof(__u64) }
> > #define PG_TBL_SZ { TARGET_PAGE_SIZE / sizeof(__u64) }
> >
> > +void rdma_dump_device_counters(Monitor *mon, RdmaDeviceResources *dev_res)
> > +{
> > + monitor_printf(mon, "\ttx : %" PRId64 "\n",
> > + dev_res->stats.tx);
> > + monitor_printf(mon, "\ttx_len : %" PRId64 "\n",
> > + dev_res->stats.tx_len);
> > + monitor_printf(mon, "\ttx_err : %" PRId64 "\n",
> > + dev_res->stats.tx_err);
> > + monitor_printf(mon, "\trx_bufs : %" PRId64 "\n",
> > + dev_res->stats.rx_bufs);
> > + monitor_printf(mon, "\trx_bufs_len : %" PRId64 "\n",
> > + dev_res->stats.rx_bufs_len);
> > + monitor_printf(mon, "\trx_bufs_err : %" PRId64 "\n",
> > + dev_res->stats.rx_bufs_err);
> > + monitor_printf(mon, "\tcomps : %" PRId64 "\n",
> > + dev_res->stats.completions);
> > + monitor_printf(mon, "\tmissing_comps : %" PRId32 "\n",
> > + dev_res->stats.missing_cqe);
> > + monitor_printf(mon, "\tpoll_cq (bk) : %" PRId64 "\n",
> > + dev_res->stats.poll_cq_from_bk);
> > + monitor_printf(mon, "\tpoll_cq_ppoll_to : %" PRId64 "\n",
> > + dev_res->stats.poll_cq_ppoll_to);
> > + monitor_printf(mon, "\tpoll_cq (fe) : %" PRId64 "\n",
> > + dev_res->stats.poll_cq_from_guest);
> > + monitor_printf(mon, "\tpoll_cq_empty : %" PRId64 "\n",
> > + dev_res->stats.poll_cq_from_guest_empty);
> > + monitor_printf(mon, "\tmad_tx : %" PRId64 "\n",
> > + dev_res->stats.mad_tx);
> > + monitor_printf(mon, "\tmad_tx_err : %" PRId64 "\n",
> > + dev_res->stats.mad_tx_err);
> > + monitor_printf(mon, "\tmad_rx : %" PRId64 "\n",
> > + dev_res->stats.mad_rx);
> > + monitor_printf(mon, "\tmad_rx_err : %" PRId64 "\n",
> > + dev_res->stats.mad_rx_err);
> > + monitor_printf(mon, "\tmad_rx_bufs : %" PRId64 "\n",
> > + dev_res->stats.mad_rx_bufs);
> > + monitor_printf(mon, "\tmad_rx_bufs_err : %" PRId64 "\n",
> > + dev_res->stats.mad_rx_bufs_err);
> > + monitor_printf(mon, "\tPDs : %" PRId32 "\n",
> > + dev_res->pd_tbl.used);
> > + monitor_printf(mon, "\tMRs : %" PRId32 "\n",
> > + dev_res->mr_tbl.used);
> > + monitor_printf(mon, "\tUCs : %" PRId32 "\n",
> > + dev_res->uc_tbl.used);
> > + monitor_printf(mon, "\tQPs : %" PRId32 "\n",
> > + dev_res->qp_tbl.used);
> > + monitor_printf(mon, "\tCQs : %" PRId32 "\n",
> > + dev_res->cq_tbl.used);
> > + monitor_printf(mon, "\tCEQ_CTXs : %" PRId32 "\n",
> > + dev_res->cqe_ctx_tbl.used);
> > +}
> > +
> > static inline void res_tbl_init(const char *name, RdmaRmResTbl *tbl,
> > uint32_t tbl_sz, uint32_t res_sz)
> > {
> > @@ -37,6 +90,7 @@ static inline void res_tbl_init(const char *name, RdmaRmResTbl *tbl,
> > tbl->bitmap = bitmap_new(tbl_sz);
> > tbl->tbl_sz = tbl_sz;
> > tbl->res_sz = res_sz;
> > + tbl->used = 0;
> > qemu_mutex_init(&tbl->lock);
> > }
> >
> > @@ -76,6 +130,8 @@ static inline void *rdma_res_tbl_alloc(RdmaRmResTbl *tbl, uint32_t *handle)
> >
> > set_bit(*handle, tbl->bitmap);
> >
> > + tbl->used++;
> > +
> > qemu_mutex_unlock(&tbl->lock);
> >
> > memset(tbl->tbl + *handle * tbl->res_sz, 0, tbl->res_sz);
> > @@ -93,6 +149,7 @@ static inline void rdma_res_tbl_dealloc(RdmaRmResTbl *tbl, uint32_t handle)
> >
> > if (handle < tbl->tbl_sz) {
> > clear_bit(handle, tbl->bitmap);
> > + tbl->used--;
> > }
> >
> > qemu_mutex_unlock(&tbl->lock);
> > @@ -620,6 +677,9 @@ int rdma_rm_init(RdmaDeviceResources *dev_res, struct ibv_device_attr *dev_attr,
> >
> > qemu_mutex_init(&dev_res->lock);
> >
> > + memset(&dev_res->stats, 0, sizeof(dev_res->stats));
> > + atomic_set(&dev_res->stats.missing_cqe, 0);
> > +
> > return 0;
> > }
> >
> > diff --git a/hw/rdma/rdma_rm.h b/hw/rdma/rdma_rm.h
> > index 9ec87d2667..a527d306f6 100644
> > --- a/hw/rdma/rdma_rm.h
> > +++ b/hw/rdma/rdma_rm.h
> > @@ -20,6 +20,7 @@
> > #include "rdma_backend_defs.h"
> > #include "rdma_rm_defs.h"
> >
> > +void rdma_dump_device_counters(Monitor *mon, RdmaDeviceResources *dev_res);
> > int rdma_rm_init(RdmaDeviceResources *dev_res, struct ibv_device_attr *dev_attr,
> > Error **errp);
> > void rdma_rm_fini(RdmaDeviceResources *dev_res, RdmaBackendDev *backend_dev,
> > diff --git a/hw/rdma/rdma_rm_defs.h b/hw/rdma/rdma_rm_defs.h
> > index f0ee1f3072..4b8d704cfe 100644
> > --- a/hw/rdma/rdma_rm_defs.h
> > +++ b/hw/rdma/rdma_rm_defs.h
> > @@ -34,7 +34,9 @@
> > #define MAX_QP_INIT_RD_ATOM 16
> > #define MAX_AH 64
> >
> > -#define MAX_RM_TBL_NAME 16
> > +#define MAX_RM_TBL_NAME 16
> > +#define MAX_CONSEQ_EMPTY_POLL_CQ 4096 /* considered as error above this */
> > +
> > typedef struct RdmaRmResTbl {
> > char name[MAX_RM_TBL_NAME];
> > QemuMutex lock;
> > @@ -42,6 +44,7 @@ typedef struct RdmaRmResTbl {
> > size_t tbl_sz;
> > size_t res_sz;
> > void *tbl;
> > + uint32_t used; /* number of used entries in the table */
> > } RdmaRmResTbl;
> >
> > typedef struct RdmaRmPD {
> > @@ -96,6 +99,27 @@ typedef struct RdmaRmPort {
> > enum ibv_port_state state;
> > } RdmaRmPort;
> >
> > +typedef struct RdmaRmStats {
> > + uint64_t tx;
> > + uint64_t tx_len;
> > + uint64_t tx_err;
> > + uint64_t rx_bufs;
> > + uint64_t rx_bufs_len;
> > + uint64_t rx_bufs_err;
> > + uint64_t completions;
> > + uint64_t mad_tx;
> > + uint64_t mad_tx_err;
> > + uint64_t mad_rx;
> > + uint64_t mad_rx_err;
> > + uint64_t mad_rx_bufs;
> > + uint64_t mad_rx_bufs_err;
> > + uint64_t poll_cq_from_bk;
> > + uint64_t poll_cq_from_guest;
> > + uint64_t poll_cq_from_guest_empty;
> > + uint64_t poll_cq_ppoll_to;
> > + uint32_t missing_cqe;
> > +} RdmaRmStats;
> > +
> > typedef struct RdmaDeviceResources {
> > RdmaRmPort port;
> > RdmaRmResTbl pd_tbl;
> > @@ -106,6 +130,7 @@ typedef struct RdmaDeviceResources {
> > RdmaRmResTbl cqe_ctx_tbl;
> > GHashTable *qp_hash; /* Keeps mapping between real and emulated */
> > QemuMutex lock;
> > + RdmaRmStats stats;
> > } RdmaDeviceResources;
> >
> > #endif
> > diff --git a/hw/rdma/vmw/pvrdma.h b/hw/rdma/vmw/pvrdma.h
> > index 0879224957..167706ec2c 100644
> > --- a/hw/rdma/vmw/pvrdma.h
> > +++ b/hw/rdma/vmw/pvrdma.h
> > @@ -70,6 +70,10 @@ typedef struct DSRInfo {
> > PvrdmaRing cq;
> > } DSRInfo;
> >
> > +typedef struct PVRDMADevStats {
> > + uint64_t commands;
> > +} PVRDMADevStats;
> > +
> > typedef struct PVRDMADev {
> > PCIDevice parent_obj;
> > MemoryRegion msix;
> > @@ -89,6 +93,7 @@ typedef struct PVRDMADev {
> > CharBackend mad_chr;
> > VMXNET3State *func0;
> > Notifier shutdown_notifier;
> > + PVRDMADevStats stats;
> > } PVRDMADev;
> > #define PVRDMA_DEV(dev) OBJECT_CHECK(PVRDMADev, (dev), PVRDMA_HW_NAME)
> >
> > diff --git a/hw/rdma/vmw/pvrdma_main.c b/hw/rdma/vmw/pvrdma_main.c
> > index b6061f4b6e..659331ac93 100644
> > --- a/hw/rdma/vmw/pvrdma_main.c
> > +++ b/hw/rdma/vmw/pvrdma_main.c
> > @@ -25,6 +25,8 @@
> > #include "cpu.h"
> > #include "trace.h"
> > #include "sysemu/sysemu.h"
> > +#include "monitor/monitor.h"
> > +#include "hw/rdma/rdma_hmp.h"
> >
> > #include "../rdma_rm.h"
> > #include "../rdma_backend.h"
> > @@ -55,6 +57,18 @@ static Property pvrdma_dev_properties[] = {
> > DEFINE_PROP_END_OF_LIST(),
> > };
> >
> > +static void pvrdma_print_statistics(Monitor *mon, RdmaStatsProvider *obj)
> > +{
> > + PVRDMADev *dev = PVRDMA_DEV(obj);
> > + PCIDevice *pdev = PCI_DEVICE(dev);
> > +
> > + monitor_printf(mon, "%s, %x.%x\n", pdev->name, PCI_SLOT(pdev->devfn),
> > + PCI_FUNC(pdev->devfn));
> > + monitor_printf(mon, "\tcommands : %" PRId64 "\n",
> > + dev->stats.commands);
> > + rdma_dump_device_counters(mon, &dev->rdma_dev_res);
> > +}
> > +
> > static void free_dev_ring(PCIDevice *pci_dev, PvrdmaRing *ring,
> > void *ring_state)
> > {
> > @@ -394,6 +408,7 @@ static void pvrdma_regs_write(void *opaque, hwaddr addr, uint64_t val,
> > if (val == 0) {
> > trace_pvrdma_regs_write(addr, val, "REQUEST", "");
> > pvrdma_exec_cmd(dev);
> > + dev->stats.commands++;
> > }
> > break;
> > default:
> > @@ -612,6 +627,8 @@ static void pvrdma_realize(PCIDevice *pdev, Error **errp)
> > goto out;
> > }
> >
> > + memset(&dev->stats, 0, sizeof(dev->stats));
> > +
> > dev->shutdown_notifier.notify = pvrdma_shutdown_notifier;
> > qemu_register_shutdown_notifier(&dev->shutdown_notifier);
> >
> > @@ -631,6 +648,7 @@ static void pvrdma_class_init(ObjectClass *klass, void *data)
> > {
> > DeviceClass *dc = DEVICE_CLASS(klass);
> > PCIDeviceClass *k = PCI_DEVICE_CLASS(klass);
> > + RdmaStatsProviderClass *ir = RDMA_STATS_PROVIDER_CLASS(klass);
> >
> > k->realize = pvrdma_realize;
> > k->exit = pvrdma_exit;
> > @@ -642,6 +660,8 @@ static void pvrdma_class_init(ObjectClass *klass, void *data)
> > dc->desc = "RDMA Device";
> > dc->props = pvrdma_dev_properties;
> > set_bit(DEVICE_CATEGORY_NETWORK, dc->categories);
> > +
> > + ir->print_statistics = pvrdma_print_statistics;
> > }
> >
> > static const TypeInfo pvrdma_info = {
> > @@ -651,6 +671,7 @@ static const TypeInfo pvrdma_info = {
> > .class_init = pvrdma_class_init,
> > .interfaces = (InterfaceInfo[]) {
> > { INTERFACE_CONVENTIONAL_PCI_DEVICE },
> > + { TYPE_RDMA_STATS_PROVIDER },
> > { }
> > }
> > };
> > diff --git a/include/hw/rdma/rdma_hmp.h b/include/hw/rdma/rdma_hmp.h
> > new file mode 100644
> > index 0000000000..dd23f2bc84
> > --- /dev/null
> > +++ b/include/hw/rdma/rdma_hmp.h
> > @@ -0,0 +1,40 @@
> > +/*
> > + * RDMA device: Human Monitor interface
> > + *
> > + * Copyright (C) 2019 Oracle
> > + * Copyright (C) 2019 Red Hat Inc
> > + *
> > + * Authors:
> > + * Yuval Shaia <yuval.shaia@oracle.com>
> > + *
> > + * This work is licensed under the terms of the GNU GPL, version 2 or later.
> > + * See the COPYING file in the top-level directory.
> > + *
> > + */
> > +
> > +#ifndef RDMA_HMP_H
> > +#define RDMA_HMP_H
> > +
> > +#include "qom/object.h"
> > +
> > +#define TYPE_RDMA_STATS_PROVIDER "rdma"
> > +
> > +#define RDMA_STATS_PROVIDER_CLASS(klass) \
> > + OBJECT_CLASS_CHECK(RdmaStatsProviderClass, (klass), \
> > + TYPE_RDMA_STATS_PROVIDER)
> > +#define RDMA_STATS_PROVIDER_GET_CLASS(obj) \
> > + OBJECT_GET_CLASS(RdmaStatsProviderClass, (obj), \
> > + TYPE_RDMA_STATS_PROVIDER)
> > +#define RDMA_STATS_PROVIDER(obj) \
> > + INTERFACE_CHECK(RdmaStatsProvider, (obj), \
> > + TYPE_RDMA_STATS_PROVIDER)
> > +
> > +typedef struct RdmaStatsProvider RdmaStatsProvider;
> > +
> > +typedef struct RdmaStatsProviderClass {
> > + InterfaceClass parent;
> > +
> > + void (*print_statistics)(Monitor *mon, RdmaStatsProvider *obj);
> > +} RdmaStatsProviderClass;
> > +
> > +#endif
> > --
> > 2.17.2
> >
> >
--
Dr. David Alan Gilbert / dgilbert@redhat.com / Manchester, UK
^ permalink raw reply [flat|nested] 17+ messages in thread
* Re: [Qemu-devel] [PATCH v4 4/9] {hmp, hw/pvrdma}: Expose device internals via monitor interface
2019-03-06 12:20 ` Dr. David Alan Gilbert
@ 2019-03-06 12:46 ` Yuval Shaia
0 siblings, 0 replies; 17+ messages in thread
From: Yuval Shaia @ 2019-03-06 12:46 UTC (permalink / raw)
To: Dr. David Alan Gilbert; +Cc: marcel.apfelbaum, armbru, qemu-devel
On Wed, Mar 06, 2019 at 12:20:49PM +0000, Dr. David Alan Gilbert wrote:
> * Yuval Shaia (yuval.shaia@oracle.com) wrote:
> > On Sun, Mar 03, 2019 at 10:33:40PM +0200, Yuval Shaia wrote:
> > > Allow interrogating device internals through HMP interface.
> > > The exposed indicators can be used for troubleshooting by developers or
> > > sysadmin.
> > > There is no need to expose these attributes to a management system (e.x.
> > > libvirt) because (1) most of them are not "device-management' related
> > > info and (2) there is no guarantee the interface is stable.
> > >
> > > Signed-off-by: Yuval Shaia <yuval.shaia@oracle.com>
> > > ---
> > > hmp-commands-info.hx | 14 ++++++++
> > > hmp.c | 27 +++++++++++++++
> > > hmp.h | 1 +
> > > hw/rdma/Makefile.objs | 2 +-
> > > hw/rdma/rdma_backend.c | 70 +++++++++++++++++++++++++++++---------
> > > hw/rdma/rdma_hmp.c | 30 ++++++++++++++++
> > > hw/rdma/rdma_rm.c | 60 ++++++++++++++++++++++++++++++++
> > > hw/rdma/rdma_rm.h | 1 +
> > > hw/rdma/rdma_rm_defs.h | 27 ++++++++++++++-
> > > hw/rdma/vmw/pvrdma.h | 5 +++
> > > hw/rdma/vmw/pvrdma_main.c | 21 ++++++++++++
> > > include/hw/rdma/rdma_hmp.h | 40 ++++++++++++++++++++++
> > > 12 files changed, 279 insertions(+), 19 deletions(-)
> > > create mode 100644 hw/rdma/rdma_hmp.c
> > > create mode 100644 include/hw/rdma/rdma_hmp.h
> > >
> > > diff --git a/hmp-commands-info.hx b/hmp-commands-info.hx
> > > index cbee8b944d..c59444c461 100644
> > > --- a/hmp-commands-info.hx
> > > +++ b/hmp-commands-info.hx
> > > @@ -202,6 +202,20 @@ STEXI
> > > @item info pic
> > > @findex info pic
> > > Show PIC state.
> > > +ETEXI
> > > +
> > > + {
> > > + .name = "rdma",
> > > + .args_type = "",
> > > + .params = "",
> > > + .help = "show RDMA state",
> > > + .cmd = hmp_info_rdma,
> > > + },
> > > +
> > > +STEXI
> > > +@item info rdma
> > > +@findex info rdma
> > > +Show RDMA state.
> > > ETEXI
> > >
> > > {
> > > diff --git a/hmp.c b/hmp.c
> > > index 1e006eeb49..68511b2441 100644
> > > --- a/hmp.c
> > > +++ b/hmp.c
> > > @@ -51,6 +51,7 @@
> > > #include "qemu/error-report.h"
> > > #include "exec/ramlist.h"
> > > #include "hw/intc/intc.h"
> > > +#include "hw/rdma/rdma_hmp.h"
> > > #include "migration/snapshot.h"
> > > #include "migration/misc.h"
> > >
> > > @@ -968,6 +969,32 @@ void hmp_info_pic(Monitor *mon, const QDict *qdict)
> > > hmp_info_pic_foreach, mon);
> > > }
> > >
> > > +static int hmp_info_rdma_foreach(Object *obj, void *opaque)
> > > +{
> > > + RdmaStatsProvider *rdma;
> > > + RdmaStatsProviderClass *k;
> > > + Monitor *mon = opaque;
> > > +
> > > + if (object_dynamic_cast(obj, TYPE_RDMA_STATS_PROVIDER)) {
> > > + rdma = RDMA_STATS_PROVIDER(obj);
> > > + k = RDMA_STATS_PROVIDER_GET_CLASS(obj);
> > > + if (k->print_statistics) {
> > > + k->print_statistics(mon, rdma);
> > > + } else {
> > > + monitor_printf(mon, "RDMA statistics not available for %s.\n",
> > > + object_get_typename(obj));
> > > + }
> > > + }
> > > +
> > > + return 0;
> > > +}
> > > +
> > > +void hmp_info_rdma(Monitor *mon, const QDict *qdict)
> > > +{
> > > + object_child_foreach_recursive(object_get_root(),
> > > + hmp_info_rdma_foreach, mon);
> > > +}
> > > +
> >
> > Hi Markus and David,
> > Is this implementation is acceptable by you?
>
> I think from the HMP side I'm fine, so:
>
> Acked-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Thanks for that.
Hope that some one can review the 'counters' part of the patch.
>
> I would say that there's really two parts to this patch, all the places
> you change your code to gather the stats, and the separate pieces that
> wire it into HMP; if you had to repost for another reason I'd
> split the patch like that.
Actually that how it was in v0 but then Markus (i think) comes up with the
idea to squash them as the *only* consumer of these counters is HMP.
But in general, i do agree to this but also can understand the other side.
>
> Dave
>
> > Thanks,
> > Yuval
> >
> >
> > > void hmp_info_pci(Monitor *mon, const QDict *qdict)
> > > {
> > > PciInfoList *info_list, *info;
> > > diff --git a/hmp.h b/hmp.h
> > > index 5f1addcca2..666949afc3 100644
> > > --- a/hmp.h
> > > +++ b/hmp.h
> > > @@ -36,6 +36,7 @@ void hmp_info_spice(Monitor *mon, const QDict *qdict);
> > > void hmp_info_balloon(Monitor *mon, const QDict *qdict);
> > > void hmp_info_irq(Monitor *mon, const QDict *qdict);
> > > void hmp_info_pic(Monitor *mon, const QDict *qdict);
> > > +void hmp_info_rdma(Monitor *mon, const QDict *qdict);
> > > void hmp_info_pci(Monitor *mon, const QDict *qdict);
> > > void hmp_info_block_jobs(Monitor *mon, const QDict *qdict);
> > > void hmp_info_tpm(Monitor *mon, const QDict *qdict);
> > > diff --git a/hw/rdma/Makefile.objs b/hw/rdma/Makefile.objs
> > > index bd36cbf51c..dd59faff51 100644
> > > --- a/hw/rdma/Makefile.objs
> > > +++ b/hw/rdma/Makefile.objs
> > > @@ -1,5 +1,5 @@
> > > ifeq ($(CONFIG_PVRDMA),y)
> > > -obj-$(CONFIG_PCI) += rdma_utils.o rdma_backend.o rdma_rm.o
> > > +obj-$(CONFIG_PCI) += rdma_utils.o rdma_backend.o rdma_rm.o rdma_hmp.o
> > > obj-$(CONFIG_PCI) += vmw/pvrdma_dev_ring.o vmw/pvrdma_cmd.o \
> > > vmw/pvrdma_qp_ops.o vmw/pvrdma_main.o
> > > endif
> > > diff --git a/hw/rdma/rdma_backend.c b/hw/rdma/rdma_backend.c
> > > index 9679b842d1..bc2fefcf93 100644
> > > --- a/hw/rdma/rdma_backend.c
> > > +++ b/hw/rdma/rdma_backend.c
> > > @@ -64,9 +64,9 @@ static inline void complete_work(enum ibv_wc_status status, uint32_t vendor_err,
> > > comp_handler(ctx, &wc);
> > > }
> > >
> > > -static void rdma_poll_cq(RdmaDeviceResources *rdma_dev_res, struct ibv_cq *ibcq)
> > > +static int rdma_poll_cq(RdmaDeviceResources *rdma_dev_res, struct ibv_cq *ibcq)
> > > {
> > > - int i, ne;
> > > + int i, ne, total_ne = 0;
> > > BackendCtx *bctx;
> > > struct ibv_wc wc[2];
> > >
> > > @@ -89,12 +89,18 @@ static void rdma_poll_cq(RdmaDeviceResources *rdma_dev_res, struct ibv_cq *ibcq)
> > > rdma_rm_dealloc_cqe_ctx(rdma_dev_res, wc[i].wr_id);
> > > g_free(bctx);
> > > }
> > > + total_ne += ne;
> > > } while (ne > 0);
> > > + atomic_sub(&rdma_dev_res->stats.missing_cqe, total_ne);
> > > qemu_mutex_unlock(&rdma_dev_res->lock);
> > >
> > > if (ne < 0) {
> > > rdma_error_report("ibv_poll_cq fail, rc=%d, errno=%d", ne, errno);
> > > }
> > > +
> > > + rdma_dev_res->stats.completions += total_ne;
> > > +
> > > + return total_ne;
> > > }
> > >
> > > static void *comp_handler_thread(void *arg)
> > > @@ -122,6 +128,9 @@ static void *comp_handler_thread(void *arg)
> > > while (backend_dev->comp_thread.run) {
> > > do {
> > > rc = qemu_poll_ns(pfds, 1, THR_POLL_TO * (int64_t)SCALE_MS);
> > > + if (!rc) {
> > > + backend_dev->rdma_dev_res->stats.poll_cq_ppoll_to++;
> > > + }
> > > } while (!rc && backend_dev->comp_thread.run);
> > >
> > > if (backend_dev->comp_thread.run) {
> > > @@ -138,6 +147,7 @@ static void *comp_handler_thread(void *arg)
> > > errno);
> > > }
> > >
> > > + backend_dev->rdma_dev_res->stats.poll_cq_from_bk++;
> > > rdma_poll_cq(backend_dev->rdma_dev_res, ev_cq);
> > >
> > > ibv_ack_cq_events(ev_cq, 1);
> > > @@ -271,7 +281,13 @@ int rdma_backend_query_port(RdmaBackendDev *backend_dev,
> > >
> > > void rdma_backend_poll_cq(RdmaDeviceResources *rdma_dev_res, RdmaBackendCQ *cq)
> > > {
> > > - rdma_poll_cq(rdma_dev_res, cq->ibcq);
> > > + int polled;
> > > +
> > > + rdma_dev_res->stats.poll_cq_from_guest++;
> > > + polled = rdma_poll_cq(rdma_dev_res, cq->ibcq);
> > > + if (!polled) {
> > > + rdma_dev_res->stats.poll_cq_from_guest_empty++;
> > > + }
> > > }
> > >
> > > static GHashTable *ah_hash;
> > > @@ -333,7 +349,7 @@ static void ah_cache_init(void)
> > >
> > > static int build_host_sge_array(RdmaDeviceResources *rdma_dev_res,
> > > struct ibv_sge *dsge, struct ibv_sge *ssge,
> > > - uint8_t num_sge)
> > > + uint8_t num_sge, uint64_t *total_length)
> > > {
> > > RdmaRmMR *mr;
> > > int ssge_idx;
> > > @@ -349,6 +365,8 @@ static int build_host_sge_array(RdmaDeviceResources *rdma_dev_res,
> > > dsge->length = ssge[ssge_idx].length;
> > > dsge->lkey = rdma_backend_mr_lkey(&mr->backend_mr);
> > >
> > > + *total_length += dsge->length;
> > > +
> > > dsge++;
> > > }
> > >
> > > @@ -445,8 +463,10 @@ void rdma_backend_post_send(RdmaBackendDev *backend_dev,
> > > rc = mad_send(backend_dev, sgid_idx, sgid, sge, num_sge);
> > > if (rc) {
> > > complete_work(IBV_WC_GENERAL_ERR, VENDOR_ERR_MAD_SEND, ctx);
> > > + backend_dev->rdma_dev_res->stats.mad_tx_err++;
> > > } else {
> > > complete_work(IBV_WC_SUCCESS, 0, ctx);
> > > + backend_dev->rdma_dev_res->stats.mad_tx++;
> > > }
> > > }
> > > return;
> > > @@ -458,20 +478,21 @@ void rdma_backend_post_send(RdmaBackendDev *backend_dev,
> > > rc = rdma_rm_alloc_cqe_ctx(backend_dev->rdma_dev_res, &bctx_id, bctx);
> > > if (unlikely(rc)) {
> > > complete_work(IBV_WC_GENERAL_ERR, VENDOR_ERR_NOMEM, ctx);
> > > - goto out_free_bctx;
> > > + goto err_free_bctx;
> > > }
> > >
> > > - rc = build_host_sge_array(backend_dev->rdma_dev_res, new_sge, sge, num_sge);
> > > + rc = build_host_sge_array(backend_dev->rdma_dev_res, new_sge, sge, num_sge,
> > > + &backend_dev->rdma_dev_res->stats.tx_len);
> > > if (rc) {
> > > complete_work(IBV_WC_GENERAL_ERR, rc, ctx);
> > > - goto out_dealloc_cqe_ctx;
> > > + goto err_dealloc_cqe_ctx;
> > > }
> > >
> > > if (qp_type == IBV_QPT_UD) {
> > > wr.wr.ud.ah = create_ah(backend_dev, qp->ibpd, sgid_idx, dgid);
> > > if (!wr.wr.ud.ah) {
> > > complete_work(IBV_WC_GENERAL_ERR, VENDOR_ERR_FAIL_BACKEND, ctx);
> > > - goto out_dealloc_cqe_ctx;
> > > + goto err_dealloc_cqe_ctx;
> > > }
> > > wr.wr.ud.remote_qpn = dqpn;
> > > wr.wr.ud.remote_qkey = dqkey;
> > > @@ -488,15 +509,19 @@ void rdma_backend_post_send(RdmaBackendDev *backend_dev,
> > > rdma_error_report("ibv_post_send fail, qpn=0x%x, rc=%d, errno=%d",
> > > qp->ibqp->qp_num, rc, errno);
> > > complete_work(IBV_WC_GENERAL_ERR, VENDOR_ERR_FAIL_BACKEND, ctx);
> > > - goto out_dealloc_cqe_ctx;
> > > + goto err_dealloc_cqe_ctx;
> > > }
> > >
> > > + atomic_inc(&backend_dev->rdma_dev_res->stats.missing_cqe);
> > > + backend_dev->rdma_dev_res->stats.tx++;
> > > +
> > > return;
> > >
> > > -out_dealloc_cqe_ctx:
> > > +err_dealloc_cqe_ctx:
> > > + backend_dev->rdma_dev_res->stats.tx_err++;
> > > rdma_rm_dealloc_cqe_ctx(backend_dev->rdma_dev_res, bctx_id);
> > >
> > > -out_free_bctx:
> > > +err_free_bctx:
> > > g_free(bctx);
> > > }
> > >
> > > @@ -554,6 +579,9 @@ void rdma_backend_post_recv(RdmaBackendDev *backend_dev,
> > > rc = save_mad_recv_buffer(backend_dev, sge, num_sge, ctx);
> > > if (rc) {
> > > complete_work(IBV_WC_GENERAL_ERR, rc, ctx);
> > > + rdma_dev_res->stats.mad_rx_bufs_err++;
> > > + } else {
> > > + rdma_dev_res->stats.mad_rx_bufs++;
> > > }
> > > }
> > > return;
> > > @@ -565,13 +593,14 @@ void rdma_backend_post_recv(RdmaBackendDev *backend_dev,
> > > rc = rdma_rm_alloc_cqe_ctx(rdma_dev_res, &bctx_id, bctx);
> > > if (unlikely(rc)) {
> > > complete_work(IBV_WC_GENERAL_ERR, VENDOR_ERR_NOMEM, ctx);
> > > - goto out_free_bctx;
> > > + goto err_free_bctx;
> > > }
> > >
> > > - rc = build_host_sge_array(rdma_dev_res, new_sge, sge, num_sge);
> > > + rc = build_host_sge_array(rdma_dev_res, new_sge, sge, num_sge,
> > > + &backend_dev->rdma_dev_res->stats.rx_bufs_len);
> > > if (rc) {
> > > complete_work(IBV_WC_GENERAL_ERR, rc, ctx);
> > > - goto out_dealloc_cqe_ctx;
> > > + goto err_dealloc_cqe_ctx;
> > > }
> > >
> > > wr.num_sge = num_sge;
> > > @@ -582,15 +611,19 @@ void rdma_backend_post_recv(RdmaBackendDev *backend_dev,
> > > rdma_error_report("ibv_post_recv fail, qpn=0x%x, rc=%d, errno=%d",
> > > qp->ibqp->qp_num, rc, errno);
> > > complete_work(IBV_WC_GENERAL_ERR, VENDOR_ERR_FAIL_BACKEND, ctx);
> > > - goto out_dealloc_cqe_ctx;
> > > + goto err_dealloc_cqe_ctx;
> > > }
> > >
> > > + atomic_inc(&backend_dev->rdma_dev_res->stats.missing_cqe);
> > > + rdma_dev_res->stats.rx_bufs++;
> > > +
> > > return;
> > >
> > > -out_dealloc_cqe_ctx:
> > > +err_dealloc_cqe_ctx:
> > > + backend_dev->rdma_dev_res->stats.rx_bufs_err++;
> > > rdma_rm_dealloc_cqe_ctx(rdma_dev_res, bctx_id);
> > >
> > > -out_free_bctx:
> > > +err_free_bctx:
> > > g_free(bctx);
> > > }
> > >
> > > @@ -929,12 +962,14 @@ static void process_incoming_mad_req(RdmaBackendDev *backend_dev,
> > > bctx = rdma_rm_get_cqe_ctx(backend_dev->rdma_dev_res, cqe_ctx_id);
> > > if (unlikely(!bctx)) {
> > > rdma_error_report("No matching ctx for req %ld", cqe_ctx_id);
> > > + backend_dev->rdma_dev_res->stats.mad_rx_err++;
> > > return;
> > > }
> > >
> > > mad = rdma_pci_dma_map(backend_dev->dev, bctx->sge.addr,
> > > bctx->sge.length);
> > > if (!mad || bctx->sge.length < msg->umad_len + MAD_HDR_SIZE) {
> > > + backend_dev->rdma_dev_res->stats.mad_rx_err++;
> > > complete_work(IBV_WC_GENERAL_ERR, VENDOR_ERR_INV_MAD_BUFF,
> > > bctx->up_ctx);
> > > } else {
> > > @@ -949,6 +984,7 @@ static void process_incoming_mad_req(RdmaBackendDev *backend_dev,
> > > wc.byte_len = msg->umad_len;
> > > wc.status = IBV_WC_SUCCESS;
> > > wc.wc_flags = IBV_WC_GRH;
> > > + backend_dev->rdma_dev_res->stats.mad_rx++;
> > > comp_handler(bctx->up_ctx, &wc);
> > > }
> > >
> > > diff --git a/hw/rdma/rdma_hmp.c b/hw/rdma/rdma_hmp.c
> > > new file mode 100644
> > > index 0000000000..c5814473c5
> > > --- /dev/null
> > > +++ b/hw/rdma/rdma_hmp.c
> > > @@ -0,0 +1,30 @@
> > > +/*
> > > + * RDMA device: Human Monitor interface
> > > + *
> > > + * Copyright (C) 2018 Oracle
> > > + * Copyright (C) 2018 Red Hat Inc
> > > + *
> > > + * Authors:
> > > + * Yuval Shaia <yuval.shaia@oracle.com>
> > > + *
> > > + * This work is licensed under the terms of the GNU GPL, version 2 or later.
> > > + * See the COPYING file in the top-level directory.
> > > + *
> > > + */
> > > +
> > > +#include "qemu/osdep.h"
> > > +#include "hw/rdma/rdma_hmp.h"
> > > +#include "qemu/module.h"
> > > +
> > > +static const TypeInfo rdma_hmp_info = {
> > > + .name = TYPE_RDMA_STATS_PROVIDER,
> > > + .parent = TYPE_INTERFACE,
> > > + .class_size = sizeof(RdmaStatsProviderClass),
> > > +};
> > > +
> > > +static void rdma_hmp_register_types(void)
> > > +{
> > > + type_register_static(&rdma_hmp_info);
> > > +}
> > > +
> > > +type_init(rdma_hmp_register_types)
> > > diff --git a/hw/rdma/rdma_rm.c b/hw/rdma/rdma_rm.c
> > > index 14580ca379..e019de1a14 100644
> > > --- a/hw/rdma/rdma_rm.c
> > > +++ b/hw/rdma/rdma_rm.c
> > > @@ -16,6 +16,7 @@
> > > #include "qemu/osdep.h"
> > > #include "qapi/error.h"
> > > #include "cpu.h"
> > > +#include "monitor/monitor.h"
> > >
> > > #include "trace.h"
> > > #include "rdma_utils.h"
> > > @@ -26,6 +27,58 @@
> > > #define PG_DIR_SZ { TARGET_PAGE_SIZE / sizeof(__u64) }
> > > #define PG_TBL_SZ { TARGET_PAGE_SIZE / sizeof(__u64) }
> > >
> > > +void rdma_dump_device_counters(Monitor *mon, RdmaDeviceResources *dev_res)
> > > +{
> > > + monitor_printf(mon, "\ttx : %" PRId64 "\n",
> > > + dev_res->stats.tx);
> > > + monitor_printf(mon, "\ttx_len : %" PRId64 "\n",
> > > + dev_res->stats.tx_len);
> > > + monitor_printf(mon, "\ttx_err : %" PRId64 "\n",
> > > + dev_res->stats.tx_err);
> > > + monitor_printf(mon, "\trx_bufs : %" PRId64 "\n",
> > > + dev_res->stats.rx_bufs);
> > > + monitor_printf(mon, "\trx_bufs_len : %" PRId64 "\n",
> > > + dev_res->stats.rx_bufs_len);
> > > + monitor_printf(mon, "\trx_bufs_err : %" PRId64 "\n",
> > > + dev_res->stats.rx_bufs_err);
> > > + monitor_printf(mon, "\tcomps : %" PRId64 "\n",
> > > + dev_res->stats.completions);
> > > + monitor_printf(mon, "\tmissing_comps : %" PRId32 "\n",
> > > + dev_res->stats.missing_cqe);
> > > + monitor_printf(mon, "\tpoll_cq (bk) : %" PRId64 "\n",
> > > + dev_res->stats.poll_cq_from_bk);
> > > + monitor_printf(mon, "\tpoll_cq_ppoll_to : %" PRId64 "\n",
> > > + dev_res->stats.poll_cq_ppoll_to);
> > > + monitor_printf(mon, "\tpoll_cq (fe) : %" PRId64 "\n",
> > > + dev_res->stats.poll_cq_from_guest);
> > > + monitor_printf(mon, "\tpoll_cq_empty : %" PRId64 "\n",
> > > + dev_res->stats.poll_cq_from_guest_empty);
> > > + monitor_printf(mon, "\tmad_tx : %" PRId64 "\n",
> > > + dev_res->stats.mad_tx);
> > > + monitor_printf(mon, "\tmad_tx_err : %" PRId64 "\n",
> > > + dev_res->stats.mad_tx_err);
> > > + monitor_printf(mon, "\tmad_rx : %" PRId64 "\n",
> > > + dev_res->stats.mad_rx);
> > > + monitor_printf(mon, "\tmad_rx_err : %" PRId64 "\n",
> > > + dev_res->stats.mad_rx_err);
> > > + monitor_printf(mon, "\tmad_rx_bufs : %" PRId64 "\n",
> > > + dev_res->stats.mad_rx_bufs);
> > > + monitor_printf(mon, "\tmad_rx_bufs_err : %" PRId64 "\n",
> > > + dev_res->stats.mad_rx_bufs_err);
> > > + monitor_printf(mon, "\tPDs : %" PRId32 "\n",
> > > + dev_res->pd_tbl.used);
> > > + monitor_printf(mon, "\tMRs : %" PRId32 "\n",
> > > + dev_res->mr_tbl.used);
> > > + monitor_printf(mon, "\tUCs : %" PRId32 "\n",
> > > + dev_res->uc_tbl.used);
> > > + monitor_printf(mon, "\tQPs : %" PRId32 "\n",
> > > + dev_res->qp_tbl.used);
> > > + monitor_printf(mon, "\tCQs : %" PRId32 "\n",
> > > + dev_res->cq_tbl.used);
> > > + monitor_printf(mon, "\tCEQ_CTXs : %" PRId32 "\n",
> > > + dev_res->cqe_ctx_tbl.used);
> > > +}
> > > +
> > > static inline void res_tbl_init(const char *name, RdmaRmResTbl *tbl,
> > > uint32_t tbl_sz, uint32_t res_sz)
> > > {
> > > @@ -37,6 +90,7 @@ static inline void res_tbl_init(const char *name, RdmaRmResTbl *tbl,
> > > tbl->bitmap = bitmap_new(tbl_sz);
> > > tbl->tbl_sz = tbl_sz;
> > > tbl->res_sz = res_sz;
> > > + tbl->used = 0;
> > > qemu_mutex_init(&tbl->lock);
> > > }
> > >
> > > @@ -76,6 +130,8 @@ static inline void *rdma_res_tbl_alloc(RdmaRmResTbl *tbl, uint32_t *handle)
> > >
> > > set_bit(*handle, tbl->bitmap);
> > >
> > > + tbl->used++;
> > > +
> > > qemu_mutex_unlock(&tbl->lock);
> > >
> > > memset(tbl->tbl + *handle * tbl->res_sz, 0, tbl->res_sz);
> > > @@ -93,6 +149,7 @@ static inline void rdma_res_tbl_dealloc(RdmaRmResTbl *tbl, uint32_t handle)
> > >
> > > if (handle < tbl->tbl_sz) {
> > > clear_bit(handle, tbl->bitmap);
> > > + tbl->used--;
> > > }
> > >
> > > qemu_mutex_unlock(&tbl->lock);
> > > @@ -620,6 +677,9 @@ int rdma_rm_init(RdmaDeviceResources *dev_res, struct ibv_device_attr *dev_attr,
> > >
> > > qemu_mutex_init(&dev_res->lock);
> > >
> > > + memset(&dev_res->stats, 0, sizeof(dev_res->stats));
> > > + atomic_set(&dev_res->stats.missing_cqe, 0);
> > > +
> > > return 0;
> > > }
> > >
> > > diff --git a/hw/rdma/rdma_rm.h b/hw/rdma/rdma_rm.h
> > > index 9ec87d2667..a527d306f6 100644
> > > --- a/hw/rdma/rdma_rm.h
> > > +++ b/hw/rdma/rdma_rm.h
> > > @@ -20,6 +20,7 @@
> > > #include "rdma_backend_defs.h"
> > > #include "rdma_rm_defs.h"
> > >
> > > +void rdma_dump_device_counters(Monitor *mon, RdmaDeviceResources *dev_res);
> > > int rdma_rm_init(RdmaDeviceResources *dev_res, struct ibv_device_attr *dev_attr,
> > > Error **errp);
> > > void rdma_rm_fini(RdmaDeviceResources *dev_res, RdmaBackendDev *backend_dev,
> > > diff --git a/hw/rdma/rdma_rm_defs.h b/hw/rdma/rdma_rm_defs.h
> > > index f0ee1f3072..4b8d704cfe 100644
> > > --- a/hw/rdma/rdma_rm_defs.h
> > > +++ b/hw/rdma/rdma_rm_defs.h
> > > @@ -34,7 +34,9 @@
> > > #define MAX_QP_INIT_RD_ATOM 16
> > > #define MAX_AH 64
> > >
> > > -#define MAX_RM_TBL_NAME 16
> > > +#define MAX_RM_TBL_NAME 16
> > > +#define MAX_CONSEQ_EMPTY_POLL_CQ 4096 /* considered as error above this */
> > > +
> > > typedef struct RdmaRmResTbl {
> > > char name[MAX_RM_TBL_NAME];
> > > QemuMutex lock;
> > > @@ -42,6 +44,7 @@ typedef struct RdmaRmResTbl {
> > > size_t tbl_sz;
> > > size_t res_sz;
> > > void *tbl;
> > > + uint32_t used; /* number of used entries in the table */
> > > } RdmaRmResTbl;
> > >
> > > typedef struct RdmaRmPD {
> > > @@ -96,6 +99,27 @@ typedef struct RdmaRmPort {
> > > enum ibv_port_state state;
> > > } RdmaRmPort;
> > >
> > > +typedef struct RdmaRmStats {
> > > + uint64_t tx;
> > > + uint64_t tx_len;
> > > + uint64_t tx_err;
> > > + uint64_t rx_bufs;
> > > + uint64_t rx_bufs_len;
> > > + uint64_t rx_bufs_err;
> > > + uint64_t completions;
> > > + uint64_t mad_tx;
> > > + uint64_t mad_tx_err;
> > > + uint64_t mad_rx;
> > > + uint64_t mad_rx_err;
> > > + uint64_t mad_rx_bufs;
> > > + uint64_t mad_rx_bufs_err;
> > > + uint64_t poll_cq_from_bk;
> > > + uint64_t poll_cq_from_guest;
> > > + uint64_t poll_cq_from_guest_empty;
> > > + uint64_t poll_cq_ppoll_to;
> > > + uint32_t missing_cqe;
> > > +} RdmaRmStats;
> > > +
> > > typedef struct RdmaDeviceResources {
> > > RdmaRmPort port;
> > > RdmaRmResTbl pd_tbl;
> > > @@ -106,6 +130,7 @@ typedef struct RdmaDeviceResources {
> > > RdmaRmResTbl cqe_ctx_tbl;
> > > GHashTable *qp_hash; /* Keeps mapping between real and emulated */
> > > QemuMutex lock;
> > > + RdmaRmStats stats;
> > > } RdmaDeviceResources;
> > >
> > > #endif
> > > diff --git a/hw/rdma/vmw/pvrdma.h b/hw/rdma/vmw/pvrdma.h
> > > index 0879224957..167706ec2c 100644
> > > --- a/hw/rdma/vmw/pvrdma.h
> > > +++ b/hw/rdma/vmw/pvrdma.h
> > > @@ -70,6 +70,10 @@ typedef struct DSRInfo {
> > > PvrdmaRing cq;
> > > } DSRInfo;
> > >
> > > +typedef struct PVRDMADevStats {
> > > + uint64_t commands;
> > > +} PVRDMADevStats;
> > > +
> > > typedef struct PVRDMADev {
> > > PCIDevice parent_obj;
> > > MemoryRegion msix;
> > > @@ -89,6 +93,7 @@ typedef struct PVRDMADev {
> > > CharBackend mad_chr;
> > > VMXNET3State *func0;
> > > Notifier shutdown_notifier;
> > > + PVRDMADevStats stats;
> > > } PVRDMADev;
> > > #define PVRDMA_DEV(dev) OBJECT_CHECK(PVRDMADev, (dev), PVRDMA_HW_NAME)
> > >
> > > diff --git a/hw/rdma/vmw/pvrdma_main.c b/hw/rdma/vmw/pvrdma_main.c
> > > index b6061f4b6e..659331ac93 100644
> > > --- a/hw/rdma/vmw/pvrdma_main.c
> > > +++ b/hw/rdma/vmw/pvrdma_main.c
> > > @@ -25,6 +25,8 @@
> > > #include "cpu.h"
> > > #include "trace.h"
> > > #include "sysemu/sysemu.h"
> > > +#include "monitor/monitor.h"
> > > +#include "hw/rdma/rdma_hmp.h"
> > >
> > > #include "../rdma_rm.h"
> > > #include "../rdma_backend.h"
> > > @@ -55,6 +57,18 @@ static Property pvrdma_dev_properties[] = {
> > > DEFINE_PROP_END_OF_LIST(),
> > > };
> > >
> > > +static void pvrdma_print_statistics(Monitor *mon, RdmaStatsProvider *obj)
> > > +{
> > > + PVRDMADev *dev = PVRDMA_DEV(obj);
> > > + PCIDevice *pdev = PCI_DEVICE(dev);
> > > +
> > > + monitor_printf(mon, "%s, %x.%x\n", pdev->name, PCI_SLOT(pdev->devfn),
> > > + PCI_FUNC(pdev->devfn));
> > > + monitor_printf(mon, "\tcommands : %" PRId64 "\n",
> > > + dev->stats.commands);
> > > + rdma_dump_device_counters(mon, &dev->rdma_dev_res);
> > > +}
> > > +
> > > static void free_dev_ring(PCIDevice *pci_dev, PvrdmaRing *ring,
> > > void *ring_state)
> > > {
> > > @@ -394,6 +408,7 @@ static void pvrdma_regs_write(void *opaque, hwaddr addr, uint64_t val,
> > > if (val == 0) {
> > > trace_pvrdma_regs_write(addr, val, "REQUEST", "");
> > > pvrdma_exec_cmd(dev);
> > > + dev->stats.commands++;
> > > }
> > > break;
> > > default:
> > > @@ -612,6 +627,8 @@ static void pvrdma_realize(PCIDevice *pdev, Error **errp)
> > > goto out;
> > > }
> > >
> > > + memset(&dev->stats, 0, sizeof(dev->stats));
> > > +
> > > dev->shutdown_notifier.notify = pvrdma_shutdown_notifier;
> > > qemu_register_shutdown_notifier(&dev->shutdown_notifier);
> > >
> > > @@ -631,6 +648,7 @@ static void pvrdma_class_init(ObjectClass *klass, void *data)
> > > {
> > > DeviceClass *dc = DEVICE_CLASS(klass);
> > > PCIDeviceClass *k = PCI_DEVICE_CLASS(klass);
> > > + RdmaStatsProviderClass *ir = RDMA_STATS_PROVIDER_CLASS(klass);
> > >
> > > k->realize = pvrdma_realize;
> > > k->exit = pvrdma_exit;
> > > @@ -642,6 +660,8 @@ static void pvrdma_class_init(ObjectClass *klass, void *data)
> > > dc->desc = "RDMA Device";
> > > dc->props = pvrdma_dev_properties;
> > > set_bit(DEVICE_CATEGORY_NETWORK, dc->categories);
> > > +
> > > + ir->print_statistics = pvrdma_print_statistics;
> > > }
> > >
> > > static const TypeInfo pvrdma_info = {
> > > @@ -651,6 +671,7 @@ static const TypeInfo pvrdma_info = {
> > > .class_init = pvrdma_class_init,
> > > .interfaces = (InterfaceInfo[]) {
> > > { INTERFACE_CONVENTIONAL_PCI_DEVICE },
> > > + { TYPE_RDMA_STATS_PROVIDER },
> > > { }
> > > }
> > > };
> > > diff --git a/include/hw/rdma/rdma_hmp.h b/include/hw/rdma/rdma_hmp.h
> > > new file mode 100644
> > > index 0000000000..dd23f2bc84
> > > --- /dev/null
> > > +++ b/include/hw/rdma/rdma_hmp.h
> > > @@ -0,0 +1,40 @@
> > > +/*
> > > + * RDMA device: Human Monitor interface
> > > + *
> > > + * Copyright (C) 2019 Oracle
> > > + * Copyright (C) 2019 Red Hat Inc
> > > + *
> > > + * Authors:
> > > + * Yuval Shaia <yuval.shaia@oracle.com>
> > > + *
> > > + * This work is licensed under the terms of the GNU GPL, version 2 or later.
> > > + * See the COPYING file in the top-level directory.
> > > + *
> > > + */
> > > +
> > > +#ifndef RDMA_HMP_H
> > > +#define RDMA_HMP_H
> > > +
> > > +#include "qom/object.h"
> > > +
> > > +#define TYPE_RDMA_STATS_PROVIDER "rdma"
> > > +
> > > +#define RDMA_STATS_PROVIDER_CLASS(klass) \
> > > + OBJECT_CLASS_CHECK(RdmaStatsProviderClass, (klass), \
> > > + TYPE_RDMA_STATS_PROVIDER)
> > > +#define RDMA_STATS_PROVIDER_GET_CLASS(obj) \
> > > + OBJECT_GET_CLASS(RdmaStatsProviderClass, (obj), \
> > > + TYPE_RDMA_STATS_PROVIDER)
> > > +#define RDMA_STATS_PROVIDER(obj) \
> > > + INTERFACE_CHECK(RdmaStatsProvider, (obj), \
> > > + TYPE_RDMA_STATS_PROVIDER)
> > > +
> > > +typedef struct RdmaStatsProvider RdmaStatsProvider;
> > > +
> > > +typedef struct RdmaStatsProviderClass {
> > > + InterfaceClass parent;
> > > +
> > > + void (*print_statistics)(Monitor *mon, RdmaStatsProvider *obj);
> > > +} RdmaStatsProviderClass;
> > > +
> > > +#endif
> > > --
> > > 2.17.2
> > >
> > >
> --
> Dr. David Alan Gilbert / dgilbert@redhat.com / Manchester, UK
^ permalink raw reply [flat|nested] 17+ messages in thread
* Re: [Qemu-devel] [PATCH v4 4/9] {hmp, hw/pvrdma}: Expose device internals via monitor interface
2019-03-03 20:33 ` [Qemu-devel] [PATCH v4 4/9] {hmp, hw/pvrdma}: Expose device internals via monitor interface Yuval Shaia
2019-03-06 10:22 ` Yuval Shaia
@ 2019-03-07 9:50 ` Marcel Apfelbaum
2019-03-08 16:37 ` Markus Armbruster
2 siblings, 0 replies; 17+ messages in thread
From: Marcel Apfelbaum @ 2019-03-07 9:50 UTC (permalink / raw)
To: Yuval Shaia, dgilbert, armbru, qemu-devel
Hi Yuval,
On 3/3/19 10:33 PM, Yuval Shaia wrote:
> Allow interrogating device internals through HMP interface.
> The exposed indicators can be used for troubleshooting by developers or
> sysadmin.
> There is no need to expose these attributes to a management system (e.x.
> libvirt) because (1) most of them are not "device-management' related
> info and (2) there is no guarantee the interface is stable.
>
> Signed-off-by: Yuval Shaia <yuval.shaia@oracle.com>
> ---
> hmp-commands-info.hx | 14 ++++++++
> hmp.c | 27 +++++++++++++++
> hmp.h | 1 +
> hw/rdma/Makefile.objs | 2 +-
> hw/rdma/rdma_backend.c | 70 +++++++++++++++++++++++++++++---------
> hw/rdma/rdma_hmp.c | 30 ++++++++++++++++
> hw/rdma/rdma_rm.c | 60 ++++++++++++++++++++++++++++++++
> hw/rdma/rdma_rm.h | 1 +
> hw/rdma/rdma_rm_defs.h | 27 ++++++++++++++-
> hw/rdma/vmw/pvrdma.h | 5 +++
> hw/rdma/vmw/pvrdma_main.c | 21 ++++++++++++
> include/hw/rdma/rdma_hmp.h | 40 ++++++++++++++++++++++
> 12 files changed, 279 insertions(+), 19 deletions(-)
> create mode 100644 hw/rdma/rdma_hmp.c
> create mode 100644 include/hw/rdma/rdma_hmp.h
>
> diff --git a/hmp-commands-info.hx b/hmp-commands-info.hx
> index cbee8b944d..c59444c461 100644
> --- a/hmp-commands-info.hx
> +++ b/hmp-commands-info.hx
> @@ -202,6 +202,20 @@ STEXI
> @item info pic
> @findex info pic
> Show PIC state.
> +ETEXI
> +
> + {
> + .name = "rdma",
> + .args_type = "",
> + .params = "",
> + .help = "show RDMA state",
> + .cmd = hmp_info_rdma,
> + },
> +
> +STEXI
> +@item info rdma
> +@findex info rdma
> +Show RDMA state.
> ETEXI
>
> {
> diff --git a/hmp.c b/hmp.c
> index 1e006eeb49..68511b2441 100644
> --- a/hmp.c
> +++ b/hmp.c
> @@ -51,6 +51,7 @@
> #include "qemu/error-report.h"
> #include "exec/ramlist.h"
> #include "hw/intc/intc.h"
> +#include "hw/rdma/rdma_hmp.h"
> #include "migration/snapshot.h"
> #include "migration/misc.h"
>
> @@ -968,6 +969,32 @@ void hmp_info_pic(Monitor *mon, const QDict *qdict)
> hmp_info_pic_foreach, mon);
> }
>
> +static int hmp_info_rdma_foreach(Object *obj, void *opaque)
> +{
> + RdmaStatsProvider *rdma;
> + RdmaStatsProviderClass *k;
> + Monitor *mon = opaque;
> +
> + if (object_dynamic_cast(obj, TYPE_RDMA_STATS_PROVIDER)) {
> + rdma = RDMA_STATS_PROVIDER(obj);
> + k = RDMA_STATS_PROVIDER_GET_CLASS(obj);
> + if (k->print_statistics) {
> + k->print_statistics(mon, rdma);
> + } else {
> + monitor_printf(mon, "RDMA statistics not available for %s.\n",
> + object_get_typename(obj));
> + }
> + }
> +
> + return 0;
> +}
> +
> +void hmp_info_rdma(Monitor *mon, const QDict *qdict)
> +{
> + object_child_foreach_recursive(object_get_root(),
> + hmp_info_rdma_foreach, mon);
> +}
> +
> void hmp_info_pci(Monitor *mon, const QDict *qdict)
> {
> PciInfoList *info_list, *info;
> diff --git a/hmp.h b/hmp.h
> index 5f1addcca2..666949afc3 100644
> --- a/hmp.h
> +++ b/hmp.h
> @@ -36,6 +36,7 @@ void hmp_info_spice(Monitor *mon, const QDict *qdict);
> void hmp_info_balloon(Monitor *mon, const QDict *qdict);
> void hmp_info_irq(Monitor *mon, const QDict *qdict);
> void hmp_info_pic(Monitor *mon, const QDict *qdict);
> +void hmp_info_rdma(Monitor *mon, const QDict *qdict);
> void hmp_info_pci(Monitor *mon, const QDict *qdict);
> void hmp_info_block_jobs(Monitor *mon, const QDict *qdict);
> void hmp_info_tpm(Monitor *mon, const QDict *qdict);
> diff --git a/hw/rdma/Makefile.objs b/hw/rdma/Makefile.objs
> index bd36cbf51c..dd59faff51 100644
> --- a/hw/rdma/Makefile.objs
> +++ b/hw/rdma/Makefile.objs
> @@ -1,5 +1,5 @@
> ifeq ($(CONFIG_PVRDMA),y)
> -obj-$(CONFIG_PCI) += rdma_utils.o rdma_backend.o rdma_rm.o
> +obj-$(CONFIG_PCI) += rdma_utils.o rdma_backend.o rdma_rm.o rdma_hmp.o
> obj-$(CONFIG_PCI) += vmw/pvrdma_dev_ring.o vmw/pvrdma_cmd.o \
> vmw/pvrdma_qp_ops.o vmw/pvrdma_main.o
> endif
> diff --git a/hw/rdma/rdma_backend.c b/hw/rdma/rdma_backend.c
> index 9679b842d1..bc2fefcf93 100644
> --- a/hw/rdma/rdma_backend.c
> +++ b/hw/rdma/rdma_backend.c
> @@ -64,9 +64,9 @@ static inline void complete_work(enum ibv_wc_status status, uint32_t vendor_err,
> comp_handler(ctx, &wc);
> }
>
> -static void rdma_poll_cq(RdmaDeviceResources *rdma_dev_res, struct ibv_cq *ibcq)
> +static int rdma_poll_cq(RdmaDeviceResources *rdma_dev_res, struct ibv_cq *ibcq)
> {
> - int i, ne;
> + int i, ne, total_ne = 0;
> BackendCtx *bctx;
> struct ibv_wc wc[2];
>
> @@ -89,12 +89,18 @@ static void rdma_poll_cq(RdmaDeviceResources *rdma_dev_res, struct ibv_cq *ibcq)
> rdma_rm_dealloc_cqe_ctx(rdma_dev_res, wc[i].wr_id);
> g_free(bctx);
> }
> + total_ne += ne;
> } while (ne > 0);
> + atomic_sub(&rdma_dev_res->stats.missing_cqe, total_ne);
> qemu_mutex_unlock(&rdma_dev_res->lock);
>
> if (ne < 0) {
> rdma_error_report("ibv_poll_cq fail, rc=%d, errno=%d", ne, errno);
> }
> +
> + rdma_dev_res->stats.completions += total_ne;
> +
> + return total_ne;
> }
>
> static void *comp_handler_thread(void *arg)
> @@ -122,6 +128,9 @@ static void *comp_handler_thread(void *arg)
> while (backend_dev->comp_thread.run) {
> do {
> rc = qemu_poll_ns(pfds, 1, THR_POLL_TO * (int64_t)SCALE_MS);
> + if (!rc) {
> + backend_dev->rdma_dev_res->stats.poll_cq_ppoll_to++;
> + }
> } while (!rc && backend_dev->comp_thread.run);
>
> if (backend_dev->comp_thread.run) {
> @@ -138,6 +147,7 @@ static void *comp_handler_thread(void *arg)
> errno);
> }
>
> + backend_dev->rdma_dev_res->stats.poll_cq_from_bk++;
> rdma_poll_cq(backend_dev->rdma_dev_res, ev_cq);
>
> ibv_ack_cq_events(ev_cq, 1);
> @@ -271,7 +281,13 @@ int rdma_backend_query_port(RdmaBackendDev *backend_dev,
>
> void rdma_backend_poll_cq(RdmaDeviceResources *rdma_dev_res, RdmaBackendCQ *cq)
> {
> - rdma_poll_cq(rdma_dev_res, cq->ibcq);
> + int polled;
> +
> + rdma_dev_res->stats.poll_cq_from_guest++;
> + polled = rdma_poll_cq(rdma_dev_res, cq->ibcq);
> + if (!polled) {
> + rdma_dev_res->stats.poll_cq_from_guest_empty++;
> + }
> }
>
> static GHashTable *ah_hash;
> @@ -333,7 +349,7 @@ static void ah_cache_init(void)
>
> static int build_host_sge_array(RdmaDeviceResources *rdma_dev_res,
> struct ibv_sge *dsge, struct ibv_sge *ssge,
> - uint8_t num_sge)
> + uint8_t num_sge, uint64_t *total_length)
> {
> RdmaRmMR *mr;
> int ssge_idx;
> @@ -349,6 +365,8 @@ static int build_host_sge_array(RdmaDeviceResources *rdma_dev_res,
> dsge->length = ssge[ssge_idx].length;
> dsge->lkey = rdma_backend_mr_lkey(&mr->backend_mr);
>
> + *total_length += dsge->length;
> +
> dsge++;
> }
>
> @@ -445,8 +463,10 @@ void rdma_backend_post_send(RdmaBackendDev *backend_dev,
> rc = mad_send(backend_dev, sgid_idx, sgid, sge, num_sge);
> if (rc) {
> complete_work(IBV_WC_GENERAL_ERR, VENDOR_ERR_MAD_SEND, ctx);
> + backend_dev->rdma_dev_res->stats.mad_tx_err++;
> } else {
> complete_work(IBV_WC_SUCCESS, 0, ctx);
> + backend_dev->rdma_dev_res->stats.mad_tx++;
> }
> }
> return;
> @@ -458,20 +478,21 @@ void rdma_backend_post_send(RdmaBackendDev *backend_dev,
> rc = rdma_rm_alloc_cqe_ctx(backend_dev->rdma_dev_res, &bctx_id, bctx);
> if (unlikely(rc)) {
> complete_work(IBV_WC_GENERAL_ERR, VENDOR_ERR_NOMEM, ctx);
> - goto out_free_bctx;
> + goto err_free_bctx;
> }
>
> - rc = build_host_sge_array(backend_dev->rdma_dev_res, new_sge, sge, num_sge);
> + rc = build_host_sge_array(backend_dev->rdma_dev_res, new_sge, sge, num_sge,
> + &backend_dev->rdma_dev_res->stats.tx_len);
> if (rc) {
> complete_work(IBV_WC_GENERAL_ERR, rc, ctx);
> - goto out_dealloc_cqe_ctx;
> + goto err_dealloc_cqe_ctx;
> }
>
> if (qp_type == IBV_QPT_UD) {
> wr.wr.ud.ah = create_ah(backend_dev, qp->ibpd, sgid_idx, dgid);
> if (!wr.wr.ud.ah) {
> complete_work(IBV_WC_GENERAL_ERR, VENDOR_ERR_FAIL_BACKEND, ctx);
> - goto out_dealloc_cqe_ctx;
> + goto err_dealloc_cqe_ctx;
> }
> wr.wr.ud.remote_qpn = dqpn;
> wr.wr.ud.remote_qkey = dqkey;
> @@ -488,15 +509,19 @@ void rdma_backend_post_send(RdmaBackendDev *backend_dev,
> rdma_error_report("ibv_post_send fail, qpn=0x%x, rc=%d, errno=%d",
> qp->ibqp->qp_num, rc, errno);
> complete_work(IBV_WC_GENERAL_ERR, VENDOR_ERR_FAIL_BACKEND, ctx);
> - goto out_dealloc_cqe_ctx;
> + goto err_dealloc_cqe_ctx;
> }
>
> + atomic_inc(&backend_dev->rdma_dev_res->stats.missing_cqe);
> + backend_dev->rdma_dev_res->stats.tx++;
> +
> return;
>
> -out_dealloc_cqe_ctx:
> +err_dealloc_cqe_ctx:
> + backend_dev->rdma_dev_res->stats.tx_err++;
> rdma_rm_dealloc_cqe_ctx(backend_dev->rdma_dev_res, bctx_id);
>
> -out_free_bctx:
> +err_free_bctx:
> g_free(bctx);
> }
>
> @@ -554,6 +579,9 @@ void rdma_backend_post_recv(RdmaBackendDev *backend_dev,
> rc = save_mad_recv_buffer(backend_dev, sge, num_sge, ctx);
> if (rc) {
> complete_work(IBV_WC_GENERAL_ERR, rc, ctx);
> + rdma_dev_res->stats.mad_rx_bufs_err++;
> + } else {
> + rdma_dev_res->stats.mad_rx_bufs++;
> }
> }
> return;
> @@ -565,13 +593,14 @@ void rdma_backend_post_recv(RdmaBackendDev *backend_dev,
> rc = rdma_rm_alloc_cqe_ctx(rdma_dev_res, &bctx_id, bctx);
> if (unlikely(rc)) {
> complete_work(IBV_WC_GENERAL_ERR, VENDOR_ERR_NOMEM, ctx);
> - goto out_free_bctx;
> + goto err_free_bctx;
> }
>
> - rc = build_host_sge_array(rdma_dev_res, new_sge, sge, num_sge);
> + rc = build_host_sge_array(rdma_dev_res, new_sge, sge, num_sge,
> + &backend_dev->rdma_dev_res->stats.rx_bufs_len);
> if (rc) {
> complete_work(IBV_WC_GENERAL_ERR, rc, ctx);
> - goto out_dealloc_cqe_ctx;
> + goto err_dealloc_cqe_ctx;
> }
>
> wr.num_sge = num_sge;
> @@ -582,15 +611,19 @@ void rdma_backend_post_recv(RdmaBackendDev *backend_dev,
> rdma_error_report("ibv_post_recv fail, qpn=0x%x, rc=%d, errno=%d",
> qp->ibqp->qp_num, rc, errno);
> complete_work(IBV_WC_GENERAL_ERR, VENDOR_ERR_FAIL_BACKEND, ctx);
> - goto out_dealloc_cqe_ctx;
> + goto err_dealloc_cqe_ctx;
> }
>
> + atomic_inc(&backend_dev->rdma_dev_res->stats.missing_cqe);
> + rdma_dev_res->stats.rx_bufs++;
> +
> return;
>
> -out_dealloc_cqe_ctx:
> +err_dealloc_cqe_ctx:
> + backend_dev->rdma_dev_res->stats.rx_bufs_err++;
> rdma_rm_dealloc_cqe_ctx(rdma_dev_res, bctx_id);
>
> -out_free_bctx:
> +err_free_bctx:
> g_free(bctx);
> }
>
> @@ -929,12 +962,14 @@ static void process_incoming_mad_req(RdmaBackendDev *backend_dev,
> bctx = rdma_rm_get_cqe_ctx(backend_dev->rdma_dev_res, cqe_ctx_id);
> if (unlikely(!bctx)) {
> rdma_error_report("No matching ctx for req %ld", cqe_ctx_id);
> + backend_dev->rdma_dev_res->stats.mad_rx_err++;
> return;
> }
>
> mad = rdma_pci_dma_map(backend_dev->dev, bctx->sge.addr,
> bctx->sge.length);
> if (!mad || bctx->sge.length < msg->umad_len + MAD_HDR_SIZE) {
> + backend_dev->rdma_dev_res->stats.mad_rx_err++;
> complete_work(IBV_WC_GENERAL_ERR, VENDOR_ERR_INV_MAD_BUFF,
> bctx->up_ctx);
> } else {
> @@ -949,6 +984,7 @@ static void process_incoming_mad_req(RdmaBackendDev *backend_dev,
> wc.byte_len = msg->umad_len;
> wc.status = IBV_WC_SUCCESS;
> wc.wc_flags = IBV_WC_GRH;
> + backend_dev->rdma_dev_res->stats.mad_rx++;
> comp_handler(bctx->up_ctx, &wc);
> }
>
> diff --git a/hw/rdma/rdma_hmp.c b/hw/rdma/rdma_hmp.c
> new file mode 100644
> index 0000000000..c5814473c5
> --- /dev/null
> +++ b/hw/rdma/rdma_hmp.c
> @@ -0,0 +1,30 @@
> +/*
> + * RDMA device: Human Monitor interface
> + *
> + * Copyright (C) 2018 Oracle
> + * Copyright (C) 2018 Red Hat Inc
> + *
> + * Authors:
> + * Yuval Shaia <yuval.shaia@oracle.com>
> + *
> + * This work is licensed under the terms of the GNU GPL, version 2 or later.
> + * See the COPYING file in the top-level directory.
> + *
> + */
> +
> +#include "qemu/osdep.h"
> +#include "hw/rdma/rdma_hmp.h"
> +#include "qemu/module.h"
> +
> +static const TypeInfo rdma_hmp_info = {
> + .name = TYPE_RDMA_STATS_PROVIDER,
> + .parent = TYPE_INTERFACE,
> + .class_size = sizeof(RdmaStatsProviderClass),
> +};
> +
> +static void rdma_hmp_register_types(void)
> +{
> + type_register_static(&rdma_hmp_info);
> +}
> +
> +type_init(rdma_hmp_register_types)
> diff --git a/hw/rdma/rdma_rm.c b/hw/rdma/rdma_rm.c
> index 14580ca379..e019de1a14 100644
> --- a/hw/rdma/rdma_rm.c
> +++ b/hw/rdma/rdma_rm.c
> @@ -16,6 +16,7 @@
> #include "qemu/osdep.h"
> #include "qapi/error.h"
> #include "cpu.h"
> +#include "monitor/monitor.h"
>
> #include "trace.h"
> #include "rdma_utils.h"
> @@ -26,6 +27,58 @@
> #define PG_DIR_SZ { TARGET_PAGE_SIZE / sizeof(__u64) }
> #define PG_TBL_SZ { TARGET_PAGE_SIZE / sizeof(__u64) }
>
> +void rdma_dump_device_counters(Monitor *mon, RdmaDeviceResources *dev_res)
> +{
> + monitor_printf(mon, "\ttx : %" PRId64 "\n",
> + dev_res->stats.tx);
> + monitor_printf(mon, "\ttx_len : %" PRId64 "\n",
> + dev_res->stats.tx_len);
> + monitor_printf(mon, "\ttx_err : %" PRId64 "\n",
> + dev_res->stats.tx_err);
> + monitor_printf(mon, "\trx_bufs : %" PRId64 "\n",
> + dev_res->stats.rx_bufs);
> + monitor_printf(mon, "\trx_bufs_len : %" PRId64 "\n",
> + dev_res->stats.rx_bufs_len);
> + monitor_printf(mon, "\trx_bufs_err : %" PRId64 "\n",
> + dev_res->stats.rx_bufs_err);
> + monitor_printf(mon, "\tcomps : %" PRId64 "\n",
> + dev_res->stats.completions);
> + monitor_printf(mon, "\tmissing_comps : %" PRId32 "\n",
> + dev_res->stats.missing_cqe);
> + monitor_printf(mon, "\tpoll_cq (bk) : %" PRId64 "\n",
> + dev_res->stats.poll_cq_from_bk);
> + monitor_printf(mon, "\tpoll_cq_ppoll_to : %" PRId64 "\n",
> + dev_res->stats.poll_cq_ppoll_to);
> + monitor_printf(mon, "\tpoll_cq (fe) : %" PRId64 "\n",
> + dev_res->stats.poll_cq_from_guest);
> + monitor_printf(mon, "\tpoll_cq_empty : %" PRId64 "\n",
> + dev_res->stats.poll_cq_from_guest_empty);
> + monitor_printf(mon, "\tmad_tx : %" PRId64 "\n",
> + dev_res->stats.mad_tx);
> + monitor_printf(mon, "\tmad_tx_err : %" PRId64 "\n",
> + dev_res->stats.mad_tx_err);
> + monitor_printf(mon, "\tmad_rx : %" PRId64 "\n",
> + dev_res->stats.mad_rx);
> + monitor_printf(mon, "\tmad_rx_err : %" PRId64 "\n",
> + dev_res->stats.mad_rx_err);
> + monitor_printf(mon, "\tmad_rx_bufs : %" PRId64 "\n",
> + dev_res->stats.mad_rx_bufs);
> + monitor_printf(mon, "\tmad_rx_bufs_err : %" PRId64 "\n",
> + dev_res->stats.mad_rx_bufs_err);
> + monitor_printf(mon, "\tPDs : %" PRId32 "\n",
> + dev_res->pd_tbl.used);
> + monitor_printf(mon, "\tMRs : %" PRId32 "\n",
> + dev_res->mr_tbl.used);
> + monitor_printf(mon, "\tUCs : %" PRId32 "\n",
> + dev_res->uc_tbl.used);
> + monitor_printf(mon, "\tQPs : %" PRId32 "\n",
> + dev_res->qp_tbl.used);
> + monitor_printf(mon, "\tCQs : %" PRId32 "\n",
> + dev_res->cq_tbl.used);
> + monitor_printf(mon, "\tCEQ_CTXs : %" PRId32 "\n",
> + dev_res->cqe_ctx_tbl.used);
> +}
> +
> static inline void res_tbl_init(const char *name, RdmaRmResTbl *tbl,
> uint32_t tbl_sz, uint32_t res_sz)
> {
> @@ -37,6 +90,7 @@ static inline void res_tbl_init(const char *name, RdmaRmResTbl *tbl,
> tbl->bitmap = bitmap_new(tbl_sz);
> tbl->tbl_sz = tbl_sz;
> tbl->res_sz = res_sz;
> + tbl->used = 0;
> qemu_mutex_init(&tbl->lock);
> }
>
> @@ -76,6 +130,8 @@ static inline void *rdma_res_tbl_alloc(RdmaRmResTbl *tbl, uint32_t *handle)
>
> set_bit(*handle, tbl->bitmap);
>
> + tbl->used++;
> +
> qemu_mutex_unlock(&tbl->lock);
>
> memset(tbl->tbl + *handle * tbl->res_sz, 0, tbl->res_sz);
> @@ -93,6 +149,7 @@ static inline void rdma_res_tbl_dealloc(RdmaRmResTbl *tbl, uint32_t handle)
>
> if (handle < tbl->tbl_sz) {
> clear_bit(handle, tbl->bitmap);
> + tbl->used--;
> }
>
> qemu_mutex_unlock(&tbl->lock);
> @@ -620,6 +677,9 @@ int rdma_rm_init(RdmaDeviceResources *dev_res, struct ibv_device_attr *dev_attr,
>
> qemu_mutex_init(&dev_res->lock);
>
> + memset(&dev_res->stats, 0, sizeof(dev_res->stats));
> + atomic_set(&dev_res->stats.missing_cqe, 0);
> +
> return 0;
> }
>
> diff --git a/hw/rdma/rdma_rm.h b/hw/rdma/rdma_rm.h
> index 9ec87d2667..a527d306f6 100644
> --- a/hw/rdma/rdma_rm.h
> +++ b/hw/rdma/rdma_rm.h
> @@ -20,6 +20,7 @@
> #include "rdma_backend_defs.h"
> #include "rdma_rm_defs.h"
>
> +void rdma_dump_device_counters(Monitor *mon, RdmaDeviceResources *dev_res);
> int rdma_rm_init(RdmaDeviceResources *dev_res, struct ibv_device_attr *dev_attr,
> Error **errp);
> void rdma_rm_fini(RdmaDeviceResources *dev_res, RdmaBackendDev *backend_dev,
> diff --git a/hw/rdma/rdma_rm_defs.h b/hw/rdma/rdma_rm_defs.h
> index f0ee1f3072..4b8d704cfe 100644
> --- a/hw/rdma/rdma_rm_defs.h
> +++ b/hw/rdma/rdma_rm_defs.h
> @@ -34,7 +34,9 @@
> #define MAX_QP_INIT_RD_ATOM 16
> #define MAX_AH 64
>
> -#define MAX_RM_TBL_NAME 16
> +#define MAX_RM_TBL_NAME 16
> +#define MAX_CONSEQ_EMPTY_POLL_CQ 4096 /* considered as error above this */
> +
> typedef struct RdmaRmResTbl {
> char name[MAX_RM_TBL_NAME];
> QemuMutex lock;
> @@ -42,6 +44,7 @@ typedef struct RdmaRmResTbl {
> size_t tbl_sz;
> size_t res_sz;
> void *tbl;
> + uint32_t used; /* number of used entries in the table */
> } RdmaRmResTbl;
>
> typedef struct RdmaRmPD {
> @@ -96,6 +99,27 @@ typedef struct RdmaRmPort {
> enum ibv_port_state state;
> } RdmaRmPort;
>
> +typedef struct RdmaRmStats {
> + uint64_t tx;
> + uint64_t tx_len;
> + uint64_t tx_err;
> + uint64_t rx_bufs;
> + uint64_t rx_bufs_len;
> + uint64_t rx_bufs_err;
> + uint64_t completions;
> + uint64_t mad_tx;
> + uint64_t mad_tx_err;
> + uint64_t mad_rx;
> + uint64_t mad_rx_err;
> + uint64_t mad_rx_bufs;
> + uint64_t mad_rx_bufs_err;
> + uint64_t poll_cq_from_bk;
> + uint64_t poll_cq_from_guest;
> + uint64_t poll_cq_from_guest_empty;
> + uint64_t poll_cq_ppoll_to;
> + uint32_t missing_cqe;
> +} RdmaRmStats;
> +
> typedef struct RdmaDeviceResources {
> RdmaRmPort port;
> RdmaRmResTbl pd_tbl;
> @@ -106,6 +130,7 @@ typedef struct RdmaDeviceResources {
> RdmaRmResTbl cqe_ctx_tbl;
> GHashTable *qp_hash; /* Keeps mapping between real and emulated */
> QemuMutex lock;
> + RdmaRmStats stats;
> } RdmaDeviceResources;
>
> #endif
> diff --git a/hw/rdma/vmw/pvrdma.h b/hw/rdma/vmw/pvrdma.h
> index 0879224957..167706ec2c 100644
> --- a/hw/rdma/vmw/pvrdma.h
> +++ b/hw/rdma/vmw/pvrdma.h
> @@ -70,6 +70,10 @@ typedef struct DSRInfo {
> PvrdmaRing cq;
> } DSRInfo;
>
> +typedef struct PVRDMADevStats {
> + uint64_t commands;
> +} PVRDMADevStats;
> +
> typedef struct PVRDMADev {
> PCIDevice parent_obj;
> MemoryRegion msix;
> @@ -89,6 +93,7 @@ typedef struct PVRDMADev {
> CharBackend mad_chr;
> VMXNET3State *func0;
> Notifier shutdown_notifier;
> + PVRDMADevStats stats;
> } PVRDMADev;
> #define PVRDMA_DEV(dev) OBJECT_CHECK(PVRDMADev, (dev), PVRDMA_HW_NAME)
>
> diff --git a/hw/rdma/vmw/pvrdma_main.c b/hw/rdma/vmw/pvrdma_main.c
> index b6061f4b6e..659331ac93 100644
> --- a/hw/rdma/vmw/pvrdma_main.c
> +++ b/hw/rdma/vmw/pvrdma_main.c
> @@ -25,6 +25,8 @@
> #include "cpu.h"
> #include "trace.h"
> #include "sysemu/sysemu.h"
> +#include "monitor/monitor.h"
> +#include "hw/rdma/rdma_hmp.h"
>
> #include "../rdma_rm.h"
> #include "../rdma_backend.h"
> @@ -55,6 +57,18 @@ static Property pvrdma_dev_properties[] = {
> DEFINE_PROP_END_OF_LIST(),
> };
>
> +static void pvrdma_print_statistics(Monitor *mon, RdmaStatsProvider *obj)
> +{
> + PVRDMADev *dev = PVRDMA_DEV(obj);
> + PCIDevice *pdev = PCI_DEVICE(dev);
> +
> + monitor_printf(mon, "%s, %x.%x\n", pdev->name, PCI_SLOT(pdev->devfn),
> + PCI_FUNC(pdev->devfn));
> + monitor_printf(mon, "\tcommands : %" PRId64 "\n",
> + dev->stats.commands);
> + rdma_dump_device_counters(mon, &dev->rdma_dev_res);
> +}
> +
> static void free_dev_ring(PCIDevice *pci_dev, PvrdmaRing *ring,
> void *ring_state)
> {
> @@ -394,6 +408,7 @@ static void pvrdma_regs_write(void *opaque, hwaddr addr, uint64_t val,
> if (val == 0) {
> trace_pvrdma_regs_write(addr, val, "REQUEST", "");
> pvrdma_exec_cmd(dev);
> + dev->stats.commands++;
> }
> break;
> default:
> @@ -612,6 +627,8 @@ static void pvrdma_realize(PCIDevice *pdev, Error **errp)
> goto out;
> }
>
> + memset(&dev->stats, 0, sizeof(dev->stats));
> +
> dev->shutdown_notifier.notify = pvrdma_shutdown_notifier;
> qemu_register_shutdown_notifier(&dev->shutdown_notifier);
>
> @@ -631,6 +648,7 @@ static void pvrdma_class_init(ObjectClass *klass, void *data)
> {
> DeviceClass *dc = DEVICE_CLASS(klass);
> PCIDeviceClass *k = PCI_DEVICE_CLASS(klass);
> + RdmaStatsProviderClass *ir = RDMA_STATS_PROVIDER_CLASS(klass);
>
> k->realize = pvrdma_realize;
> k->exit = pvrdma_exit;
> @@ -642,6 +660,8 @@ static void pvrdma_class_init(ObjectClass *klass, void *data)
> dc->desc = "RDMA Device";
> dc->props = pvrdma_dev_properties;
> set_bit(DEVICE_CATEGORY_NETWORK, dc->categories);
> +
> + ir->print_statistics = pvrdma_print_statistics;
> }
>
> static const TypeInfo pvrdma_info = {
> @@ -651,6 +671,7 @@ static const TypeInfo pvrdma_info = {
> .class_init = pvrdma_class_init,
> .interfaces = (InterfaceInfo[]) {
> { INTERFACE_CONVENTIONAL_PCI_DEVICE },
> + { TYPE_RDMA_STATS_PROVIDER },
> { }
> }
> };
> diff --git a/include/hw/rdma/rdma_hmp.h b/include/hw/rdma/rdma_hmp.h
> new file mode 100644
> index 0000000000..dd23f2bc84
> --- /dev/null
> +++ b/include/hw/rdma/rdma_hmp.h
> @@ -0,0 +1,40 @@
> +/*
> + * RDMA device: Human Monitor interface
> + *
> + * Copyright (C) 2019 Oracle
> + * Copyright (C) 2019 Red Hat Inc
> + *
> + * Authors:
> + * Yuval Shaia <yuval.shaia@oracle.com>
> + *
> + * This work is licensed under the terms of the GNU GPL, version 2 or later.
> + * See the COPYING file in the top-level directory.
> + *
> + */
> +
> +#ifndef RDMA_HMP_H
> +#define RDMA_HMP_H
> +
> +#include "qom/object.h"
> +
> +#define TYPE_RDMA_STATS_PROVIDER "rdma"
> +
> +#define RDMA_STATS_PROVIDER_CLASS(klass) \
> + OBJECT_CLASS_CHECK(RdmaStatsProviderClass, (klass), \
> + TYPE_RDMA_STATS_PROVIDER)
> +#define RDMA_STATS_PROVIDER_GET_CLASS(obj) \
> + OBJECT_GET_CLASS(RdmaStatsProviderClass, (obj), \
> + TYPE_RDMA_STATS_PROVIDER)
> +#define RDMA_STATS_PROVIDER(obj) \
> + INTERFACE_CHECK(RdmaStatsProvider, (obj), \
> + TYPE_RDMA_STATS_PROVIDER)
> +
> +typedef struct RdmaStatsProvider RdmaStatsProvider;
> +
> +typedef struct RdmaStatsProviderClass {
> + InterfaceClass parent;
> +
> + void (*print_statistics)(Monitor *mon, RdmaStatsProvider *obj);
> +} RdmaStatsProviderClass;
> +
> +#endif
RDMAStats implementation looks ok to me.
Reviewed-by: Marcel Apfelbaum<marcel.apfelbaum@gmail.com>
Thanks,
Marcel
^ permalink raw reply [flat|nested] 17+ messages in thread
* Re: [Qemu-devel] [PATCH v4 4/9] {hmp, hw/pvrdma}: Expose device internals via monitor interface
2019-03-03 20:33 ` [Qemu-devel] [PATCH v4 4/9] {hmp, hw/pvrdma}: Expose device internals via monitor interface Yuval Shaia
2019-03-06 10:22 ` Yuval Shaia
2019-03-07 9:50 ` Marcel Apfelbaum
@ 2019-03-08 16:37 ` Markus Armbruster
2019-03-08 18:57 ` Yuval Shaia
2019-03-10 8:06 ` Yuval Shaia
2 siblings, 2 replies; 17+ messages in thread
From: Markus Armbruster @ 2019-03-08 16:37 UTC (permalink / raw)
To: Yuval Shaia; +Cc: dgilbert, marcel.apfelbaum, qemu-devel
Yuval Shaia <yuval.shaia@oracle.com> writes:
> Allow interrogating device internals through HMP interface.
> The exposed indicators can be used for troubleshooting by developers or
> sysadmin.
> There is no need to expose these attributes to a management system (e.x.
> libvirt) because (1) most of them are not "device-management' related
> info and (2) there is no guarantee the interface is stable.
>
> Signed-off-by: Yuval Shaia <yuval.shaia@oracle.com>
> ---
> hmp-commands-info.hx | 14 ++++++++
> hmp.c | 27 +++++++++++++++
> hmp.h | 1 +
> hw/rdma/Makefile.objs | 2 +-
> hw/rdma/rdma_backend.c | 70 +++++++++++++++++++++++++++++---------
> hw/rdma/rdma_hmp.c | 30 ++++++++++++++++
> hw/rdma/rdma_rm.c | 60 ++++++++++++++++++++++++++++++++
> hw/rdma/rdma_rm.h | 1 +
> hw/rdma/rdma_rm_defs.h | 27 ++++++++++++++-
> hw/rdma/vmw/pvrdma.h | 5 +++
> hw/rdma/vmw/pvrdma_main.c | 21 ++++++++++++
> include/hw/rdma/rdma_hmp.h | 40 ++++++++++++++++++++++
> 12 files changed, 279 insertions(+), 19 deletions(-)
> create mode 100644 hw/rdma/rdma_hmp.c
> create mode 100644 include/hw/rdma/rdma_hmp.h
>
> diff --git a/hmp-commands-info.hx b/hmp-commands-info.hx
> index cbee8b944d..c59444c461 100644
> --- a/hmp-commands-info.hx
> +++ b/hmp-commands-info.hx
> @@ -202,6 +202,20 @@ STEXI
> @item info pic
> @findex info pic
> Show PIC state.
> +ETEXI
> +
> + {
> + .name = "rdma",
> + .args_type = "",
> + .params = "",
> + .help = "show RDMA state",
> + .cmd = hmp_info_rdma,
> + },
> +
> +STEXI
> +@item info rdma
> +@findex info rdma
> +Show RDMA state.
> ETEXI
>
> {
> diff --git a/hmp.c b/hmp.c
> index 1e006eeb49..68511b2441 100644
> --- a/hmp.c
> +++ b/hmp.c
> @@ -51,6 +51,7 @@
> #include "qemu/error-report.h"
> #include "exec/ramlist.h"
> #include "hw/intc/intc.h"
> +#include "hw/rdma/rdma_hmp.h"
> #include "migration/snapshot.h"
> #include "migration/misc.h"
>
> @@ -968,6 +969,32 @@ void hmp_info_pic(Monitor *mon, const QDict *qdict)
> hmp_info_pic_foreach, mon);
> }
>
> +static int hmp_info_rdma_foreach(Object *obj, void *opaque)
> +{
> + RdmaStatsProvider *rdma;
> + RdmaStatsProviderClass *k;
> + Monitor *mon = opaque;
> +
> + if (object_dynamic_cast(obj, TYPE_RDMA_STATS_PROVIDER)) {
> + rdma = RDMA_STATS_PROVIDER(obj);
> + k = RDMA_STATS_PROVIDER_GET_CLASS(obj);
> + if (k->print_statistics) {
> + k->print_statistics(mon, rdma);
> + } else {
> + monitor_printf(mon, "RDMA statistics not available for %s.\n",
> + object_get_typename(obj));
> + }
> + }
> +
> + return 0;
> +}
> +
> +void hmp_info_rdma(Monitor *mon, const QDict *qdict)
> +{
> + object_child_foreach_recursive(object_get_root(),
> + hmp_info_rdma_foreach, mon);
> +}
> +
> void hmp_info_pci(Monitor *mon, const QDict *qdict)
> {
> PciInfoList *info_list, *info;
> diff --git a/hmp.h b/hmp.h
> index 5f1addcca2..666949afc3 100644
> --- a/hmp.h
> +++ b/hmp.h
> @@ -36,6 +36,7 @@ void hmp_info_spice(Monitor *mon, const QDict *qdict);
> void hmp_info_balloon(Monitor *mon, const QDict *qdict);
> void hmp_info_irq(Monitor *mon, const QDict *qdict);
> void hmp_info_pic(Monitor *mon, const QDict *qdict);
> +void hmp_info_rdma(Monitor *mon, const QDict *qdict);
> void hmp_info_pci(Monitor *mon, const QDict *qdict);
> void hmp_info_block_jobs(Monitor *mon, const QDict *qdict);
> void hmp_info_tpm(Monitor *mon, const QDict *qdict);
HMP stuff looks good to me.
> diff --git a/hw/rdma/Makefile.objs b/hw/rdma/Makefile.objs
> index bd36cbf51c..dd59faff51 100644
> --- a/hw/rdma/Makefile.objs
> +++ b/hw/rdma/Makefile.objs
> @@ -1,5 +1,5 @@
> ifeq ($(CONFIG_PVRDMA),y)
> -obj-$(CONFIG_PCI) += rdma_utils.o rdma_backend.o rdma_rm.o
> +obj-$(CONFIG_PCI) += rdma_utils.o rdma_backend.o rdma_rm.o rdma_hmp.o
> obj-$(CONFIG_PCI) += vmw/pvrdma_dev_ring.o vmw/pvrdma_cmd.o \
> vmw/pvrdma_qp_ops.o vmw/pvrdma_main.o
> endif
> diff --git a/hw/rdma/rdma_backend.c b/hw/rdma/rdma_backend.c
> index 9679b842d1..bc2fefcf93 100644
> --- a/hw/rdma/rdma_backend.c
> +++ b/hw/rdma/rdma_backend.c
The changes to this file look like they're for gathering stats. The
patch would be much simpler to review had you kept the stats gathering
separate. I can only review the HMP / QOM part, and to do that, I have
to identify what to ignore.
[...]
> diff --git a/hw/rdma/rdma_hmp.c b/hw/rdma/rdma_hmp.c
> new file mode 100644
> index 0000000000..c5814473c5
> --- /dev/null
> +++ b/hw/rdma/rdma_hmp.c
> @@ -0,0 +1,30 @@
> +/*
> + * RDMA device: Human Monitor interface
The file name and this comment are a bit akward. Yes, you create
TYPE_RDMA_STATS_PROVIDER for use in HMP info rdma, but there's
absolutely nothing HMP-related in this file. Same for rdma_hmp.h below.
Call them rdma_stats.c and rdma_stats.h?
> + *
> + * Copyright (C) 2018 Oracle
> + * Copyright (C) 2018 Red Hat Inc
> + *
> + * Authors:
> + * Yuval Shaia <yuval.shaia@oracle.com>
> + *
> + * This work is licensed under the terms of the GNU GPL, version 2 or later.
> + * See the COPYING file in the top-level directory.
> + *
> + */
> +
> +#include "qemu/osdep.h"
> +#include "hw/rdma/rdma_hmp.h"
> +#include "qemu/module.h"
> +
> +static const TypeInfo rdma_hmp_info = {
> + .name = TYPE_RDMA_STATS_PROVIDER,
> + .parent = TYPE_INTERFACE,
> + .class_size = sizeof(RdmaStatsProviderClass),
> +};
> +
> +static void rdma_hmp_register_types(void)
> +{
> + type_register_static(&rdma_hmp_info);
> +}
> +
> +type_init(rdma_hmp_register_types)
Also rename _hmp_ to _stats_.
> diff --git a/hw/rdma/rdma_rm.c b/hw/rdma/rdma_rm.c
> index 14580ca379..e019de1a14 100644
> --- a/hw/rdma/rdma_rm.c
> +++ b/hw/rdma/rdma_rm.c
> @@ -16,6 +16,7 @@
> #include "qemu/osdep.h"
> #include "qapi/error.h"
> #include "cpu.h"
> +#include "monitor/monitor.h"
>
> #include "trace.h"
> #include "rdma_utils.h"
> @@ -26,6 +27,58 @@
> #define PG_DIR_SZ { TARGET_PAGE_SIZE / sizeof(__u64) }
> #define PG_TBL_SZ { TARGET_PAGE_SIZE / sizeof(__u64) }
>
> +void rdma_dump_device_counters(Monitor *mon, RdmaDeviceResources *dev_res)
> +{
> + monitor_printf(mon, "\ttx : %" PRId64 "\n",
> + dev_res->stats.tx);
> + monitor_printf(mon, "\ttx_len : %" PRId64 "\n",
> + dev_res->stats.tx_len);
> + monitor_printf(mon, "\ttx_err : %" PRId64 "\n",
> + dev_res->stats.tx_err);
> + monitor_printf(mon, "\trx_bufs : %" PRId64 "\n",
> + dev_res->stats.rx_bufs);
> + monitor_printf(mon, "\trx_bufs_len : %" PRId64 "\n",
> + dev_res->stats.rx_bufs_len);
> + monitor_printf(mon, "\trx_bufs_err : %" PRId64 "\n",
> + dev_res->stats.rx_bufs_err);
> + monitor_printf(mon, "\tcomps : %" PRId64 "\n",
> + dev_res->stats.completions);
> + monitor_printf(mon, "\tmissing_comps : %" PRId32 "\n",
> + dev_res->stats.missing_cqe);
> + monitor_printf(mon, "\tpoll_cq (bk) : %" PRId64 "\n",
> + dev_res->stats.poll_cq_from_bk);
> + monitor_printf(mon, "\tpoll_cq_ppoll_to : %" PRId64 "\n",
> + dev_res->stats.poll_cq_ppoll_to);
> + monitor_printf(mon, "\tpoll_cq (fe) : %" PRId64 "\n",
> + dev_res->stats.poll_cq_from_guest);
> + monitor_printf(mon, "\tpoll_cq_empty : %" PRId64 "\n",
> + dev_res->stats.poll_cq_from_guest_empty);
> + monitor_printf(mon, "\tmad_tx : %" PRId64 "\n",
> + dev_res->stats.mad_tx);
> + monitor_printf(mon, "\tmad_tx_err : %" PRId64 "\n",
> + dev_res->stats.mad_tx_err);
> + monitor_printf(mon, "\tmad_rx : %" PRId64 "\n",
> + dev_res->stats.mad_rx);
> + monitor_printf(mon, "\tmad_rx_err : %" PRId64 "\n",
> + dev_res->stats.mad_rx_err);
> + monitor_printf(mon, "\tmad_rx_bufs : %" PRId64 "\n",
> + dev_res->stats.mad_rx_bufs);
> + monitor_printf(mon, "\tmad_rx_bufs_err : %" PRId64 "\n",
> + dev_res->stats.mad_rx_bufs_err);
> + monitor_printf(mon, "\tPDs : %" PRId32 "\n",
> + dev_res->pd_tbl.used);
> + monitor_printf(mon, "\tMRs : %" PRId32 "\n",
> + dev_res->mr_tbl.used);
> + monitor_printf(mon, "\tUCs : %" PRId32 "\n",
> + dev_res->uc_tbl.used);
> + monitor_printf(mon, "\tQPs : %" PRId32 "\n",
> + dev_res->qp_tbl.used);
> + monitor_printf(mon, "\tCQs : %" PRId32 "\n",
> + dev_res->cq_tbl.used);
> + monitor_printf(mon, "\tCEQ_CTXs : %" PRId32 "\n",
> + dev_res->cqe_ctx_tbl.used);
> +}
> +
> static inline void res_tbl_init(const char *name, RdmaRmResTbl *tbl,
> uint32_t tbl_sz, uint32_t res_sz)
> {
> @@ -37,6 +90,7 @@ static inline void res_tbl_init(const char *name, RdmaRmResTbl *tbl,
More stats gathering, I think.
[...]
> diff --git a/hw/rdma/rdma_rm.h b/hw/rdma/rdma_rm.h
> index 9ec87d2667..a527d306f6 100644
> --- a/hw/rdma/rdma_rm.h
> +++ b/hw/rdma/rdma_rm.h
> @@ -20,6 +20,7 @@
> #include "rdma_backend_defs.h"
> #include "rdma_rm_defs.h"
>
> +void rdma_dump_device_counters(Monitor *mon, RdmaDeviceResources *dev_res);
> int rdma_rm_init(RdmaDeviceResources *dev_res, struct ibv_device_attr *dev_attr,
> Error **errp);
> void rdma_rm_fini(RdmaDeviceResources *dev_res, RdmaBackendDev *backend_dev,
> diff --git a/hw/rdma/rdma_rm_defs.h b/hw/rdma/rdma_rm_defs.h
> index f0ee1f3072..4b8d704cfe 100644
> --- a/hw/rdma/rdma_rm_defs.h
> +++ b/hw/rdma/rdma_rm_defs.h
Still more stats gathering, I think.
[...]
> diff --git a/hw/rdma/vmw/pvrdma_main.c b/hw/rdma/vmw/pvrdma_main.c
> index b6061f4b6e..659331ac93 100644
> --- a/hw/rdma/vmw/pvrdma_main.c
> +++ b/hw/rdma/vmw/pvrdma_main.c
> @@ -25,6 +25,8 @@
> #include "cpu.h"
> #include "trace.h"
> #include "sysemu/sysemu.h"
> +#include "monitor/monitor.h"
> +#include "hw/rdma/rdma_hmp.h"
>
> #include "../rdma_rm.h"
> #include "../rdma_backend.h"
> @@ -55,6 +57,18 @@ static Property pvrdma_dev_properties[] = {
> DEFINE_PROP_END_OF_LIST(),
> };
>
> +static void pvrdma_print_statistics(Monitor *mon, RdmaStatsProvider *obj)
> +{
> + PVRDMADev *dev = PVRDMA_DEV(obj);
> + PCIDevice *pdev = PCI_DEVICE(dev);
> +
> + monitor_printf(mon, "%s, %x.%x\n", pdev->name, PCI_SLOT(pdev->devfn),
> + PCI_FUNC(pdev->devfn));
> + monitor_printf(mon, "\tcommands : %" PRId64 "\n",
> + dev->stats.commands);
> + rdma_dump_device_counters(mon, &dev->rdma_dev_res);
> +}
> +
> static void free_dev_ring(PCIDevice *pci_dev, PvrdmaRing *ring,
> void *ring_state)
> {
> @@ -394,6 +408,7 @@ static void pvrdma_regs_write(void *opaque, hwaddr addr, uint64_t val,
Still more stats gathering, I think.
[...]
> @@ -631,6 +648,7 @@ static void pvrdma_class_init(ObjectClass *klass, void *data)
> {
> DeviceClass *dc = DEVICE_CLASS(klass);
> PCIDeviceClass *k = PCI_DEVICE_CLASS(klass);
> + RdmaStatsProviderClass *ir = RDMA_STATS_PROVIDER_CLASS(klass);
>
> k->realize = pvrdma_realize;
> k->exit = pvrdma_exit;
> @@ -642,6 +660,8 @@ static void pvrdma_class_init(ObjectClass *klass, void *data)
> dc->desc = "RDMA Device";
> dc->props = pvrdma_dev_properties;
> set_bit(DEVICE_CATEGORY_NETWORK, dc->categories);
> +
> + ir->print_statistics = pvrdma_print_statistics;
> }
>
> static const TypeInfo pvrdma_info = {
> @@ -651,6 +671,7 @@ static const TypeInfo pvrdma_info = {
> .class_init = pvrdma_class_init,
> .interfaces = (InterfaceInfo[]) {
> { INTERFACE_CONVENTIONAL_PCI_DEVICE },
> + { TYPE_RDMA_STATS_PROVIDER },
We use INTERFACE_ rather than TYPE_ for some, but not all interfaces.
You follow TYPE_INTERRUPT_STATS_PROVIDER precedence. We suck. Not your
problem to solve.
> { }
> }
> };
> diff --git a/include/hw/rdma/rdma_hmp.h b/include/hw/rdma/rdma_hmp.h
> new file mode 100644
> index 0000000000..dd23f2bc84
> --- /dev/null
> +++ b/include/hw/rdma/rdma_hmp.h
> @@ -0,0 +1,40 @@
> +/*
> + * RDMA device: Human Monitor interface
> + *
> + * Copyright (C) 2019 Oracle
> + * Copyright (C) 2019 Red Hat Inc
> + *
> + * Authors:
> + * Yuval Shaia <yuval.shaia@oracle.com>
> + *
> + * This work is licensed under the terms of the GNU GPL, version 2 or later.
> + * See the COPYING file in the top-level directory.
> + *
> + */
> +
> +#ifndef RDMA_HMP_H
> +#define RDMA_HMP_H
> +
> +#include "qom/object.h"
> +
> +#define TYPE_RDMA_STATS_PROVIDER "rdma"
> +
> +#define RDMA_STATS_PROVIDER_CLASS(klass) \
> + OBJECT_CLASS_CHECK(RdmaStatsProviderClass, (klass), \
> + TYPE_RDMA_STATS_PROVIDER)
> +#define RDMA_STATS_PROVIDER_GET_CLASS(obj) \
> + OBJECT_GET_CLASS(RdmaStatsProviderClass, (obj), \
> + TYPE_RDMA_STATS_PROVIDER)
> +#define RDMA_STATS_PROVIDER(obj) \
> + INTERFACE_CHECK(RdmaStatsProvider, (obj), \
> + TYPE_RDMA_STATS_PROVIDER)
> +
> +typedef struct RdmaStatsProvider RdmaStatsProvider;
> +
> +typedef struct RdmaStatsProviderClass {
> + InterfaceClass parent;
> +
> + void (*print_statistics)(Monitor *mon, RdmaStatsProvider *obj);
> +} RdmaStatsProviderClass;
> +
> +#endif
I think using QOM to find the RDMA devices came out rather nicely.
Thanks for giving it a try.
Please consider the renames I recommended.
Acked-by: Markus Armbruster <armbru@redhat.com>
^ permalink raw reply [flat|nested] 17+ messages in thread
* Re: [Qemu-devel] [PATCH v4 4/9] {hmp, hw/pvrdma}: Expose device internals via monitor interface
2019-03-08 16:37 ` Markus Armbruster
@ 2019-03-08 18:57 ` Yuval Shaia
2019-03-10 8:06 ` Yuval Shaia
1 sibling, 0 replies; 17+ messages in thread
From: Yuval Shaia @ 2019-03-08 18:57 UTC (permalink / raw)
To: Markus Armbruster; +Cc: dgilbert, marcel.apfelbaum, qemu-devel
On Fri, Mar 08, 2019 at 05:37:45PM +0100, Markus Armbruster wrote:
> Yuval Shaia <yuval.shaia@oracle.com> writes:
>
> > Allow interrogating device internals through HMP interface.
> > The exposed indicators can be used for troubleshooting by developers or
> > sysadmin.
> > There is no need to expose these attributes to a management system (e.x.
> > libvirt) because (1) most of them are not "device-management' related
> > info and (2) there is no guarantee the interface is stable.
> >
> > Signed-off-by: Yuval Shaia <yuval.shaia@oracle.com>
> > ---
> > hmp-commands-info.hx | 14 ++++++++
> > hmp.c | 27 +++++++++++++++
> > hmp.h | 1 +
> > hw/rdma/Makefile.objs | 2 +-
> > hw/rdma/rdma_backend.c | 70 +++++++++++++++++++++++++++++---------
> > hw/rdma/rdma_hmp.c | 30 ++++++++++++++++
> > hw/rdma/rdma_rm.c | 60 ++++++++++++++++++++++++++++++++
> > hw/rdma/rdma_rm.h | 1 +
> > hw/rdma/rdma_rm_defs.h | 27 ++++++++++++++-
> > hw/rdma/vmw/pvrdma.h | 5 +++
> > hw/rdma/vmw/pvrdma_main.c | 21 ++++++++++++
> > include/hw/rdma/rdma_hmp.h | 40 ++++++++++++++++++++++
> > 12 files changed, 279 insertions(+), 19 deletions(-)
> > create mode 100644 hw/rdma/rdma_hmp.c
> > create mode 100644 include/hw/rdma/rdma_hmp.h
> >
> > diff --git a/hmp-commands-info.hx b/hmp-commands-info.hx
> > index cbee8b944d..c59444c461 100644
> > --- a/hmp-commands-info.hx
> > +++ b/hmp-commands-info.hx
> > @@ -202,6 +202,20 @@ STEXI
> > @item info pic
> > @findex info pic
> > Show PIC state.
> > +ETEXI
> > +
> > + {
> > + .name = "rdma",
> > + .args_type = "",
> > + .params = "",
> > + .help = "show RDMA state",
> > + .cmd = hmp_info_rdma,
> > + },
> > +
> > +STEXI
> > +@item info rdma
> > +@findex info rdma
> > +Show RDMA state.
> > ETEXI
> >
> > {
> > diff --git a/hmp.c b/hmp.c
> > index 1e006eeb49..68511b2441 100644
> > --- a/hmp.c
> > +++ b/hmp.c
> > @@ -51,6 +51,7 @@
> > #include "qemu/error-report.h"
> > #include "exec/ramlist.h"
> > #include "hw/intc/intc.h"
> > +#include "hw/rdma/rdma_hmp.h"
> > #include "migration/snapshot.h"
> > #include "migration/misc.h"
> >
> > @@ -968,6 +969,32 @@ void hmp_info_pic(Monitor *mon, const QDict *qdict)
> > hmp_info_pic_foreach, mon);
> > }
> >
> > +static int hmp_info_rdma_foreach(Object *obj, void *opaque)
> > +{
> > + RdmaStatsProvider *rdma;
> > + RdmaStatsProviderClass *k;
> > + Monitor *mon = opaque;
> > +
> > + if (object_dynamic_cast(obj, TYPE_RDMA_STATS_PROVIDER)) {
> > + rdma = RDMA_STATS_PROVIDER(obj);
> > + k = RDMA_STATS_PROVIDER_GET_CLASS(obj);
> > + if (k->print_statistics) {
> > + k->print_statistics(mon, rdma);
> > + } else {
> > + monitor_printf(mon, "RDMA statistics not available for %s.\n",
> > + object_get_typename(obj));
> > + }
> > + }
> > +
> > + return 0;
> > +}
> > +
> > +void hmp_info_rdma(Monitor *mon, const QDict *qdict)
> > +{
> > + object_child_foreach_recursive(object_get_root(),
> > + hmp_info_rdma_foreach, mon);
> > +}
> > +
> > void hmp_info_pci(Monitor *mon, const QDict *qdict)
> > {
> > PciInfoList *info_list, *info;
> > diff --git a/hmp.h b/hmp.h
> > index 5f1addcca2..666949afc3 100644
> > --- a/hmp.h
> > +++ b/hmp.h
> > @@ -36,6 +36,7 @@ void hmp_info_spice(Monitor *mon, const QDict *qdict);
> > void hmp_info_balloon(Monitor *mon, const QDict *qdict);
> > void hmp_info_irq(Monitor *mon, const QDict *qdict);
> > void hmp_info_pic(Monitor *mon, const QDict *qdict);
> > +void hmp_info_rdma(Monitor *mon, const QDict *qdict);
> > void hmp_info_pci(Monitor *mon, const QDict *qdict);
> > void hmp_info_block_jobs(Monitor *mon, const QDict *qdict);
> > void hmp_info_tpm(Monitor *mon, const QDict *qdict);
>
> HMP stuff looks good to me.
>
> > diff --git a/hw/rdma/Makefile.objs b/hw/rdma/Makefile.objs
> > index bd36cbf51c..dd59faff51 100644
> > --- a/hw/rdma/Makefile.objs
> > +++ b/hw/rdma/Makefile.objs
> > @@ -1,5 +1,5 @@
> > ifeq ($(CONFIG_PVRDMA),y)
> > -obj-$(CONFIG_PCI) += rdma_utils.o rdma_backend.o rdma_rm.o
> > +obj-$(CONFIG_PCI) += rdma_utils.o rdma_backend.o rdma_rm.o rdma_hmp.o
> > obj-$(CONFIG_PCI) += vmw/pvrdma_dev_ring.o vmw/pvrdma_cmd.o \
> > vmw/pvrdma_qp_ops.o vmw/pvrdma_main.o
> > endif
> > diff --git a/hw/rdma/rdma_backend.c b/hw/rdma/rdma_backend.c
> > index 9679b842d1..bc2fefcf93 100644
> > --- a/hw/rdma/rdma_backend.c
> > +++ b/hw/rdma/rdma_backend.c
>
> The changes to this file look like they're for gathering stats. The
> patch would be much simpler to review had you kept the stats gathering
> separate. I can only review the HMP / QOM part, and to do that, I have
> to identify what to ignore.
>
> [...]
> > diff --git a/hw/rdma/rdma_hmp.c b/hw/rdma/rdma_hmp.c
> > new file mode 100644
> > index 0000000000..c5814473c5
> > --- /dev/null
> > +++ b/hw/rdma/rdma_hmp.c
> > @@ -0,0 +1,30 @@
> > +/*
> > + * RDMA device: Human Monitor interface
>
> The file name and this comment are a bit akward. Yes, you create
> TYPE_RDMA_STATS_PROVIDER for use in HMP info rdma, but there's
> absolutely nothing HMP-related in this file. Same for rdma_hmp.h below.
>
> Call them rdma_stats.c and rdma_stats.h?
>
> > + *
> > + * Copyright (C) 2018 Oracle
> > + * Copyright (C) 2018 Red Hat Inc
> > + *
> > + * Authors:
> > + * Yuval Shaia <yuval.shaia@oracle.com>
> > + *
> > + * This work is licensed under the terms of the GNU GPL, version 2 or later.
> > + * See the COPYING file in the top-level directory.
> > + *
> > + */
> > +
> > +#include "qemu/osdep.h"
> > +#include "hw/rdma/rdma_hmp.h"
> > +#include "qemu/module.h"
> > +
> > +static const TypeInfo rdma_hmp_info = {
> > + .name = TYPE_RDMA_STATS_PROVIDER,
> > + .parent = TYPE_INTERFACE,
> > + .class_size = sizeof(RdmaStatsProviderClass),
> > +};
> > +
> > +static void rdma_hmp_register_types(void)
> > +{
> > + type_register_static(&rdma_hmp_info);
> > +}
> > +
> > +type_init(rdma_hmp_register_types)
>
> Also rename _hmp_ to _stats_.
>
> > diff --git a/hw/rdma/rdma_rm.c b/hw/rdma/rdma_rm.c
> > index 14580ca379..e019de1a14 100644
> > --- a/hw/rdma/rdma_rm.c
> > +++ b/hw/rdma/rdma_rm.c
> > @@ -16,6 +16,7 @@
> > #include "qemu/osdep.h"
> > #include "qapi/error.h"
> > #include "cpu.h"
> > +#include "monitor/monitor.h"
> >
> > #include "trace.h"
> > #include "rdma_utils.h"
> > @@ -26,6 +27,58 @@
> > #define PG_DIR_SZ { TARGET_PAGE_SIZE / sizeof(__u64) }
> > #define PG_TBL_SZ { TARGET_PAGE_SIZE / sizeof(__u64) }
> >
> > +void rdma_dump_device_counters(Monitor *mon, RdmaDeviceResources *dev_res)
> > +{
> > + monitor_printf(mon, "\ttx : %" PRId64 "\n",
> > + dev_res->stats.tx);
> > + monitor_printf(mon, "\ttx_len : %" PRId64 "\n",
> > + dev_res->stats.tx_len);
> > + monitor_printf(mon, "\ttx_err : %" PRId64 "\n",
> > + dev_res->stats.tx_err);
> > + monitor_printf(mon, "\trx_bufs : %" PRId64 "\n",
> > + dev_res->stats.rx_bufs);
> > + monitor_printf(mon, "\trx_bufs_len : %" PRId64 "\n",
> > + dev_res->stats.rx_bufs_len);
> > + monitor_printf(mon, "\trx_bufs_err : %" PRId64 "\n",
> > + dev_res->stats.rx_bufs_err);
> > + monitor_printf(mon, "\tcomps : %" PRId64 "\n",
> > + dev_res->stats.completions);
> > + monitor_printf(mon, "\tmissing_comps : %" PRId32 "\n",
> > + dev_res->stats.missing_cqe);
> > + monitor_printf(mon, "\tpoll_cq (bk) : %" PRId64 "\n",
> > + dev_res->stats.poll_cq_from_bk);
> > + monitor_printf(mon, "\tpoll_cq_ppoll_to : %" PRId64 "\n",
> > + dev_res->stats.poll_cq_ppoll_to);
> > + monitor_printf(mon, "\tpoll_cq (fe) : %" PRId64 "\n",
> > + dev_res->stats.poll_cq_from_guest);
> > + monitor_printf(mon, "\tpoll_cq_empty : %" PRId64 "\n",
> > + dev_res->stats.poll_cq_from_guest_empty);
> > + monitor_printf(mon, "\tmad_tx : %" PRId64 "\n",
> > + dev_res->stats.mad_tx);
> > + monitor_printf(mon, "\tmad_tx_err : %" PRId64 "\n",
> > + dev_res->stats.mad_tx_err);
> > + monitor_printf(mon, "\tmad_rx : %" PRId64 "\n",
> > + dev_res->stats.mad_rx);
> > + monitor_printf(mon, "\tmad_rx_err : %" PRId64 "\n",
> > + dev_res->stats.mad_rx_err);
> > + monitor_printf(mon, "\tmad_rx_bufs : %" PRId64 "\n",
> > + dev_res->stats.mad_rx_bufs);
> > + monitor_printf(mon, "\tmad_rx_bufs_err : %" PRId64 "\n",
> > + dev_res->stats.mad_rx_bufs_err);
> > + monitor_printf(mon, "\tPDs : %" PRId32 "\n",
> > + dev_res->pd_tbl.used);
> > + monitor_printf(mon, "\tMRs : %" PRId32 "\n",
> > + dev_res->mr_tbl.used);
> > + monitor_printf(mon, "\tUCs : %" PRId32 "\n",
> > + dev_res->uc_tbl.used);
> > + monitor_printf(mon, "\tQPs : %" PRId32 "\n",
> > + dev_res->qp_tbl.used);
> > + monitor_printf(mon, "\tCQs : %" PRId32 "\n",
> > + dev_res->cq_tbl.used);
> > + monitor_printf(mon, "\tCEQ_CTXs : %" PRId32 "\n",
> > + dev_res->cqe_ctx_tbl.used);
> > +}
> > +
> > static inline void res_tbl_init(const char *name, RdmaRmResTbl *tbl,
> > uint32_t tbl_sz, uint32_t res_sz)
> > {
> > @@ -37,6 +90,7 @@ static inline void res_tbl_init(const char *name, RdmaRmResTbl *tbl,
>
> More stats gathering, I think.
>
> [...]
> > diff --git a/hw/rdma/rdma_rm.h b/hw/rdma/rdma_rm.h
> > index 9ec87d2667..a527d306f6 100644
> > --- a/hw/rdma/rdma_rm.h
> > +++ b/hw/rdma/rdma_rm.h
> > @@ -20,6 +20,7 @@
> > #include "rdma_backend_defs.h"
> > #include "rdma_rm_defs.h"
> >
> > +void rdma_dump_device_counters(Monitor *mon, RdmaDeviceResources *dev_res);
> > int rdma_rm_init(RdmaDeviceResources *dev_res, struct ibv_device_attr *dev_attr,
> > Error **errp);
> > void rdma_rm_fini(RdmaDeviceResources *dev_res, RdmaBackendDev *backend_dev,
> > diff --git a/hw/rdma/rdma_rm_defs.h b/hw/rdma/rdma_rm_defs.h
> > index f0ee1f3072..4b8d704cfe 100644
> > --- a/hw/rdma/rdma_rm_defs.h
> > +++ b/hw/rdma/rdma_rm_defs.h
>
> Still more stats gathering, I think.
>
> [...]
> > diff --git a/hw/rdma/vmw/pvrdma_main.c b/hw/rdma/vmw/pvrdma_main.c
> > index b6061f4b6e..659331ac93 100644
> > --- a/hw/rdma/vmw/pvrdma_main.c
> > +++ b/hw/rdma/vmw/pvrdma_main.c
> > @@ -25,6 +25,8 @@
> > #include "cpu.h"
> > #include "trace.h"
> > #include "sysemu/sysemu.h"
> > +#include "monitor/monitor.h"
> > +#include "hw/rdma/rdma_hmp.h"
> >
> > #include "../rdma_rm.h"
> > #include "../rdma_backend.h"
> > @@ -55,6 +57,18 @@ static Property pvrdma_dev_properties[] = {
> > DEFINE_PROP_END_OF_LIST(),
> > };
> >
> > +static void pvrdma_print_statistics(Monitor *mon, RdmaStatsProvider *obj)
> > +{
> > + PVRDMADev *dev = PVRDMA_DEV(obj);
> > + PCIDevice *pdev = PCI_DEVICE(dev);
> > +
> > + monitor_printf(mon, "%s, %x.%x\n", pdev->name, PCI_SLOT(pdev->devfn),
> > + PCI_FUNC(pdev->devfn));
> > + monitor_printf(mon, "\tcommands : %" PRId64 "\n",
> > + dev->stats.commands);
> > + rdma_dump_device_counters(mon, &dev->rdma_dev_res);
> > +}
> > +
> > static void free_dev_ring(PCIDevice *pci_dev, PvrdmaRing *ring,
> > void *ring_state)
> > {
> > @@ -394,6 +408,7 @@ static void pvrdma_regs_write(void *opaque, hwaddr addr, uint64_t val,
>
> Still more stats gathering, I think.
>
> [...]
> > @@ -631,6 +648,7 @@ static void pvrdma_class_init(ObjectClass *klass, void *data)
> > {
> > DeviceClass *dc = DEVICE_CLASS(klass);
> > PCIDeviceClass *k = PCI_DEVICE_CLASS(klass);
> > + RdmaStatsProviderClass *ir = RDMA_STATS_PROVIDER_CLASS(klass);
> >
> > k->realize = pvrdma_realize;
> > k->exit = pvrdma_exit;
> > @@ -642,6 +660,8 @@ static void pvrdma_class_init(ObjectClass *klass, void *data)
> > dc->desc = "RDMA Device";
> > dc->props = pvrdma_dev_properties;
> > set_bit(DEVICE_CATEGORY_NETWORK, dc->categories);
> > +
> > + ir->print_statistics = pvrdma_print_statistics;
> > }
> >
> > static const TypeInfo pvrdma_info = {
> > @@ -651,6 +671,7 @@ static const TypeInfo pvrdma_info = {
> > .class_init = pvrdma_class_init,
> > .interfaces = (InterfaceInfo[]) {
> > { INTERFACE_CONVENTIONAL_PCI_DEVICE },
> > + { TYPE_RDMA_STATS_PROVIDER },
>
> We use INTERFACE_ rather than TYPE_ for some, but not all interfaces.
> You follow TYPE_INTERRUPT_STATS_PROVIDER precedence. We suck. Not your
> problem to solve.
>
> > { }
> > }
> > };
> > diff --git a/include/hw/rdma/rdma_hmp.h b/include/hw/rdma/rdma_hmp.h
> > new file mode 100644
> > index 0000000000..dd23f2bc84
> > --- /dev/null
> > +++ b/include/hw/rdma/rdma_hmp.h
> > @@ -0,0 +1,40 @@
> > +/*
> > + * RDMA device: Human Monitor interface
> > + *
> > + * Copyright (C) 2019 Oracle
> > + * Copyright (C) 2019 Red Hat Inc
> > + *
> > + * Authors:
> > + * Yuval Shaia <yuval.shaia@oracle.com>
> > + *
> > + * This work is licensed under the terms of the GNU GPL, version 2 or later.
> > + * See the COPYING file in the top-level directory.
> > + *
> > + */
> > +
> > +#ifndef RDMA_HMP_H
> > +#define RDMA_HMP_H
> > +
> > +#include "qom/object.h"
> > +
> > +#define TYPE_RDMA_STATS_PROVIDER "rdma"
> > +
> > +#define RDMA_STATS_PROVIDER_CLASS(klass) \
> > + OBJECT_CLASS_CHECK(RdmaStatsProviderClass, (klass), \
> > + TYPE_RDMA_STATS_PROVIDER)
> > +#define RDMA_STATS_PROVIDER_GET_CLASS(obj) \
> > + OBJECT_GET_CLASS(RdmaStatsProviderClass, (obj), \
> > + TYPE_RDMA_STATS_PROVIDER)
> > +#define RDMA_STATS_PROVIDER(obj) \
> > + INTERFACE_CHECK(RdmaStatsProvider, (obj), \
> > + TYPE_RDMA_STATS_PROVIDER)
> > +
> > +typedef struct RdmaStatsProvider RdmaStatsProvider;
> > +
> > +typedef struct RdmaStatsProviderClass {
> > + InterfaceClass parent;
> > +
> > + void (*print_statistics)(Monitor *mon, RdmaStatsProvider *obj);
> > +} RdmaStatsProviderClass;
> > +
> > +#endif
>
> I think using QOM to find the RDMA devices came out rather nicely.
> Thanks for giving it a try.
>
> Please consider the renames I recommended.
Will do, it perfectly make sense.
>
> Acked-by: Markus Armbruster <armbru@redhat.com>
Thanks.
^ permalink raw reply [flat|nested] 17+ messages in thread
* Re: [Qemu-devel] [PATCH v4 4/9] {hmp, hw/pvrdma}: Expose device internals via monitor interface
2019-03-08 16:37 ` Markus Armbruster
2019-03-08 18:57 ` Yuval Shaia
@ 2019-03-10 8:06 ` Yuval Shaia
1 sibling, 0 replies; 17+ messages in thread
From: Yuval Shaia @ 2019-03-10 8:06 UTC (permalink / raw)
To: Markus Armbruster; +Cc: dgilbert, marcel.apfelbaum, qemu-devel
>
> [...]
> > diff --git a/hw/rdma/rdma_hmp.c b/hw/rdma/rdma_hmp.c
> > new file mode 100644
> > index 0000000000..c5814473c5
> > --- /dev/null
> > +++ b/hw/rdma/rdma_hmp.c
> > @@ -0,0 +1,30 @@
> > +/*
> > + * RDMA device: Human Monitor interface
>
> The file name and this comment are a bit akward. Yes, you create
> TYPE_RDMA_STATS_PROVIDER for use in HMP info rdma, but there's
> absolutely nothing HMP-related in this file. Same for rdma_hmp.h below.
>
> Call them rdma_stats.c and rdma_stats.h?
Renamed to rdma.h and rdma.c, wo knows what other things might be added in
the future.
>
> > + *
> > + * Copyright (C) 2018 Oracle
> > + * Copyright (C) 2018 Red Hat Inc
> > + *
> > + * Authors:
> > + * Yuval Shaia <yuval.shaia@oracle.com>
> > + *
> > + * This work is licensed under the terms of the GNU GPL, version 2 or later.
> > + * See the COPYING file in the top-level directory.
> > + *
> > + */
> > +
> > +#include "qemu/osdep.h"
> > +#include "hw/rdma/rdma_hmp.h"
> > +#include "qemu/module.h"
> > +
> > +static const TypeInfo rdma_hmp_info = {
> > + .name = TYPE_RDMA_STATS_PROVIDER,
> > + .parent = TYPE_INTERFACE,
> > + .class_size = sizeof(RdmaStatsProviderClass),
> > +};
> > +
> > +static void rdma_hmp_register_types(void)
> > +{
> > + type_register_static(&rdma_hmp_info);
> > +}
> > +
> > +type_init(rdma_hmp_register_types)
>
> Also rename _hmp_ to _stats_.
Ditto.
>
^ permalink raw reply [flat|nested] 17+ messages in thread
end of thread, other threads:[~2019-03-10 8:06 UTC | newest]
Thread overview: 17+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2019-03-03 20:33 [Qemu-devel] [PATCH v4 0/9] Misc fixes to pvrdma device Yuval Shaia
2019-03-03 20:33 ` [Qemu-devel] [PATCH v4 1/9] hw/rdma: Switch to generic error reporting way Yuval Shaia
2019-03-03 20:33 ` [Qemu-devel] [PATCH v4 2/9] hw/rdma: Introduce protected qlist Yuval Shaia
2019-03-03 20:33 ` [Qemu-devel] [PATCH v4 3/9] hw/rdma: Protect against concurrent execution of poll_cq Yuval Shaia
2019-03-03 20:33 ` [Qemu-devel] [PATCH v4 4/9] {hmp, hw/pvrdma}: Expose device internals via monitor interface Yuval Shaia
2019-03-06 10:22 ` Yuval Shaia
2019-03-06 12:20 ` Dr. David Alan Gilbert
2019-03-06 12:46 ` Yuval Shaia
2019-03-07 9:50 ` Marcel Apfelbaum
2019-03-08 16:37 ` Markus Armbruster
2019-03-08 18:57 ` Yuval Shaia
2019-03-10 8:06 ` Yuval Shaia
2019-03-03 20:33 ` [Qemu-devel] [PATCH v4 5/9] hw/rdma: Free all MAD receive buffers when device is closed Yuval Shaia
2019-03-03 20:33 ` [Qemu-devel] [PATCH v4 6/9] hw/rdma: Free all receive buffers when QP is destroyed Yuval Shaia
2019-03-03 20:33 ` [Qemu-devel] [PATCH v4 7/9] hw/pvrdma: Delete unneeded function argument Yuval Shaia
2019-03-03 20:33 ` [Qemu-devel] [PATCH v4 8/9] hw/pvrdma: Delete pvrdma_exit function Yuval Shaia
2019-03-03 20:33 ` [Qemu-devel] [PATCH v4 9/9] hw/pvrdma: Unregister from shutdown notifier when device goes down Yuval Shaia
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).