RDMA queue

* Clang compilation fix
  * Coverity fix
  * Various fixes for the pvrdma device
 -----BEGIN PGP SIGNATURE-----
 
 iQEcBAABAgAGBQJcQupRAAoJEDbUwPDPL+RtlUkIALWOiY0kCWUrleLBrmlP8iDs
 bsv24vQsttL/2yIjf95p5dlv9mU6WTlhC3PjgqtCti8x7Lfw5zxUrQXX4Pe9kVo/
 Zw5iaRvTtI6A+2w5av4DbBoKhXfegz8my+Q8A10gyRTa2IiFhzZf6tR5yttgZBaI
 89IpCKrMZ1PF/Si8o5JWKItKbv2nF86EgzeorzsQro5fHzbAAuFif8yCGIeev8v6
 aNCLEBYjUBs3opLGCjgwkif2s1Xypknr5/WJQ/MMInkM3pH0Q2kg7ig6RzPXGIAy
 gKV/6hGcbeE38d45GH3/q8cDrjuTno5f+OcUVbwAveuKdbiAbB2dH3tjkoVt6zs=
 =E6e1
 -----END PGP SIGNATURE-----

Merge remote-tracking branch 'remotes/marcel/tags/rdma-pull-request' into staging

RDMA queue
 * Clang compilation fix
 * Coverity fix
 * Various fixes for the pvrdma device

# gpg: Signature made Sat 19 Jan 2019 09:13:53 GMT
# gpg:                using RSA key 36D4C0F0CF2FE46D
# gpg: Good signature from "Marcel Apfelbaum <marcel.apfelbaum@zoho.com>"
# gpg:                 aka "Marcel Apfelbaum <marcel@redhat.com>"
# gpg:                 aka "Marcel Apfelbaum <marcel.apfelbaum@gmail.com>"
# gpg: WARNING: This key is not certified with sufficiently trusted signatures!
# gpg:          It is not certain that the signature belongs to the owner.
# Primary key fingerprint: B1C6 3A57 F92E 08F2 640F  31F5 36D4 C0F0 CF2F E46D

* remotes/marcel/tags/rdma-pull-request:
  contrib/rdmacm-mux: fix clang compilation
  hw/rdma: modify struct initialization
  contrib/rdmacm-mux: remove Wno-format-truncation flag
  hw: rdma: fix an off-by-one issue
  hw/rdma: Verify that ptr is not NULL before freeing
  hw/pvrdma: Make function pvrdma_qp_send/recv return void.
  hw/pvrdma: Post CQE when receive invalid gid index
  hw/rdma: Delete unused struct member
  hw/pvrdma: Remove max-sge command-line param
  docs/pvrdma: Update rdmacm-mux documentation

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
This commit is contained in:
Peter Maydell 2019-01-21 19:19:47 +00:00
commit e56b86bc77
11 changed files with 96 additions and 66 deletions

View File

@ -581,6 +581,8 @@ vhost-user-scsi$(EXESUF): $(vhost-user-scsi-obj-y) libvhost-user.a
$(call LINK, $^)
vhost-user-blk$(EXESUF): $(vhost-user-blk-obj-y) libvhost-user.a
$(call LINK, $^)
rdmacm-mux$(EXESUF): LIBS += "-libumad"
rdmacm-mux$(EXESUF): $(rdmacm-mux-obj-y) $(COMMON_LDADDS)
$(call LINK, $^)

View File

@ -1,4 +1,3 @@
ifdef CONFIG_PVRDMA
CFLAGS += -libumad -Wno-format-truncation
rdmacm-mux-obj-y = main.o
endif

View File

@ -42,6 +42,8 @@
/* The below can be override by command line parameter */
#define UNIX_SOCKET_PATH "/var/run/rdmacm-mux"
/* Has format %s-%s-%d" <path>-<rdma-dev--name>-<port> */
#define SOCKET_PATH_MAX (PATH_MAX - NAME_MAX - sizeof(int) - 2)
#define RDMA_PORT_NUM 1
typedef struct RdmaCmServerArgs {
@ -95,7 +97,7 @@ static void help(const char *progname)
static void parse_args(int argc, char *argv[])
{
int c;
char unix_socket_path[PATH_MAX];
char unix_socket_path[SOCKET_PATH_MAX];
strcpy(server.args.rdma_dev_name, "");
strcpy(unix_socket_path, UNIX_SOCKET_PATH);
@ -113,7 +115,7 @@ static void parse_args(int argc, char *argv[])
case 's':
/* This is temporary, final name will build below */
strncpy(unix_socket_path, optarg, PATH_MAX);
strncpy(unix_socket_path, optarg, SOCKET_PATH_MAX);
break;
case 'p':
@ -348,7 +350,7 @@ static int get_fd(const char *mad, int *fd, __be64 *gid_ifid)
static void *umad_recv_thread_func(void *args)
{
int rc;
RdmaCmMuxMsg msg = {0};
RdmaCmMuxMsg msg = {};
int fd = -2;
msg.hdr.msg_type = RDMACM_MUX_MSG_TYPE_REQ;
@ -385,7 +387,7 @@ static void *umad_recv_thread_func(void *args)
static int read_and_process(int fd)
{
int rc;
RdmaCmMuxMsg msg = {0};
RdmaCmMuxMsg msg = {};
struct umad_hdr *hdr;
uint32_t *comm_id = 0;
uint16_t attr_id;
@ -742,7 +744,7 @@ static void signal_handler(int sig, siginfo_t *siginfo, void *context)
static int init(void)
{
int rc;
struct sigaction sig = {0};
struct sigaction sig = {};
rc = init_listener();
if (rc) {

View File

@ -99,6 +99,9 @@ MAD layer to send and receive RDMA-CM MAD packets.
To build rdmacm-mux run
# make rdmacm-mux
Before running the rdmacm-mux make sure that both ib_cm and rdma_cm kernel
modules aren't loaded, otherwise the rdmacm-mux service will fail to start.
The application accepts 3 command line arguments and exposes a UNIX socket
to pass control and data to it.
-d rdma-device-name Name of RDMA device to register with
@ -153,7 +156,6 @@ Ethernet function can be used for other Ethernet purposes such as IP.
specify the port to use. If not set 1 will be used.
- dev-caps-max-mr-size: The maximum size of MR.
- dev-caps-max-qp: Maximum number of QPs.
- dev-caps-max-sge: Maximum number of SGE elements in WR.
- dev-caps-max-cq: Maximum number of CQs.
- dev-caps-max-mr: Maximum number of MRs.
- dev-caps-max-pd: Maximum number of PDs.

View File

@ -32,17 +32,6 @@
#include "rdma_rm.h"
#include "rdma_backend.h"
/* Vendor Errors */
#define VENDOR_ERR_FAIL_BACKEND 0x201
#define VENDOR_ERR_TOO_MANY_SGES 0x202
#define VENDOR_ERR_NOMEM 0x203
#define VENDOR_ERR_QP0 0x204
#define VENDOR_ERR_INV_NUM_SGE 0x205
#define VENDOR_ERR_MAD_SEND 0x206
#define VENDOR_ERR_INVLKEY 0x207
#define VENDOR_ERR_MR_SMALL 0x208
#define VENDOR_ERR_INV_MAD_BUFF 0x209
#define THR_NAME_LEN 16
#define THR_POLL_TO 5000
@ -190,7 +179,7 @@ static inline int rdmacm_mux_can_process_async(RdmaBackendDev *backend_dev)
static int check_mux_op_status(CharBackend *mad_chr_be)
{
RdmaCmMuxMsg msg = {0};
RdmaCmMuxMsg msg = {};
int ret;
pr_dbg("Reading response\n");
@ -387,7 +376,7 @@ static int build_host_sge_array(RdmaDeviceResources *rdma_dev_res,
static int mad_send(RdmaBackendDev *backend_dev, uint8_t sgid_idx,
union ibv_gid *sgid, struct ibv_sge *sge, uint32_t num_sge)
{
RdmaCmMuxMsg msg = {0};
RdmaCmMuxMsg msg = {};
char *hdr, *data;
int ret;
@ -475,11 +464,6 @@ void rdma_backend_post_send(RdmaBackendDev *backend_dev,
}
pr_dbg("num_sge=%d\n", num_sge);
if (!num_sge || num_sge > MAX_SGE) {
pr_dbg("invalid num_sge=%d\n", num_sge);
complete_work(IBV_WC_GENERAL_ERR, VENDOR_ERR_INV_NUM_SGE, ctx);
return;
}
bctx = g_malloc0(sizeof(*bctx));
bctx->up_ctx = ctx;
@ -602,11 +586,6 @@ void rdma_backend_post_recv(RdmaBackendDev *backend_dev,
}
pr_dbg("num_sge=%d\n", num_sge);
if (!num_sge || num_sge > MAX_SGE) {
pr_dbg("invalid num_sge=%d\n", num_sge);
complete_work(IBV_WC_GENERAL_ERR, VENDOR_ERR_INV_NUM_SGE, ctx);
return;
}
bctx = g_malloc0(sizeof(*bctx));
bctx->up_ctx = ctx;
@ -938,21 +917,25 @@ void rdma_backend_destroy_qp(RdmaBackendQP *qp)
static int init_device_caps(RdmaBackendDev *backend_dev,
struct ibv_device_attr *dev_attr)
{
if (ibv_query_device(backend_dev->context, &backend_dev->dev_attr)) {
struct ibv_device_attr bk_dev_attr;
if (ibv_query_device(backend_dev->context, &bk_dev_attr)) {
return -EIO;
}
CHK_ATTR(dev_attr, backend_dev->dev_attr, max_mr_size, "%" PRId64);
CHK_ATTR(dev_attr, backend_dev->dev_attr, max_qp, "%d");
CHK_ATTR(dev_attr, backend_dev->dev_attr, max_sge, "%d");
CHK_ATTR(dev_attr, backend_dev->dev_attr, max_qp_wr, "%d");
CHK_ATTR(dev_attr, backend_dev->dev_attr, max_cq, "%d");
CHK_ATTR(dev_attr, backend_dev->dev_attr, max_cqe, "%d");
CHK_ATTR(dev_attr, backend_dev->dev_attr, max_mr, "%d");
CHK_ATTR(dev_attr, backend_dev->dev_attr, max_pd, "%d");
CHK_ATTR(dev_attr, backend_dev->dev_attr, max_qp_rd_atom, "%d");
CHK_ATTR(dev_attr, backend_dev->dev_attr, max_qp_init_rd_atom, "%d");
CHK_ATTR(dev_attr, backend_dev->dev_attr, max_ah, "%d");
dev_attr->max_sge = MAX_SGE;
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");
CHK_ATTR(dev_attr, bk_dev_attr, max_qp_init_rd_atom, "%d");
CHK_ATTR(dev_attr, bk_dev_attr, max_ah, "%d");
return 0;
}
@ -1083,9 +1066,11 @@ 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) {
qlist_destroy_obj(QOBJECT(backend_dev->recv_mads_list.list));
qemu_mutex_destroy(&backend_dev->recv_mads_list.lock);
}
}
int rdma_backend_get_gid_index(RdmaBackendDev *backend_dev,
union ibv_gid *gid)
@ -1112,7 +1097,7 @@ int rdma_backend_get_gid_index(RdmaBackendDev *backend_dev,
int rdma_backend_add_gid(RdmaBackendDev *backend_dev, const char *ifname,
union ibv_gid *gid)
{
RdmaCmMuxMsg msg = {0};
RdmaCmMuxMsg msg = {};
int ret;
pr_dbg("0x%llx, 0x%llx\n",
@ -1138,7 +1123,7 @@ int rdma_backend_add_gid(RdmaBackendDev *backend_dev, const char *ifname,
int rdma_backend_del_gid(RdmaBackendDev *backend_dev, const char *ifname,
union ibv_gid *gid)
{
RdmaCmMuxMsg msg = {0};
RdmaCmMuxMsg msg = {};
int ret;
pr_dbg("0x%llx, 0x%llx\n",

View File

@ -22,6 +22,18 @@
#include "rdma_rm_defs.h"
#include "rdma_backend_defs.h"
/* Vendor Errors */
#define VENDOR_ERR_FAIL_BACKEND 0x201
#define VENDOR_ERR_TOO_MANY_SGES 0x202
#define VENDOR_ERR_NOMEM 0x203
#define VENDOR_ERR_QP0 0x204
#define VENDOR_ERR_INV_NUM_SGE 0x205
#define VENDOR_ERR_MAD_SEND 0x206
#define VENDOR_ERR_INVLKEY 0x207
#define VENDOR_ERR_MR_SMALL 0x208
#define VENDOR_ERR_INV_MAD_BUFF 0x209
#define VENDOR_ERR_INV_GID_IDX 0x210
/* Add definition for QP0 and QP1 as there is no userspace enums for them */
enum ibv_special_qp_type {
IBV_QPT_SMI = 0,

View File

@ -41,7 +41,6 @@ typedef struct RdmaCmMux {
} RdmaCmMux;
typedef struct RdmaBackendDev {
struct ibv_device_attr dev_attr;
RdmaBackendThread comp_thread;
PCIDevice *dev;
RdmaDeviceResources *rdma_dev_res;

View File

@ -41,6 +41,9 @@ static inline void res_tbl_init(const char *name, RdmaRmResTbl *tbl,
static inline void res_tbl_free(RdmaRmResTbl *tbl)
{
if (!tbl->bitmap) {
return;
}
qemu_mutex_destroy(&tbl->lock);
g_free(tbl->tbl);
g_free(tbl->bitmap);
@ -576,7 +579,7 @@ int rdma_rm_del_gid(RdmaDeviceResources *dev_res, RdmaBackendDev *backend_dev,
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)) {
if (unlikely(sgid_idx < 0 || sgid_idx >= MAX_PORT_GIDS)) {
pr_dbg("Got invalid sgid_idx %d\n", sgid_idx);
return -EINVAL;
}
@ -655,5 +658,7 @@ void rdma_rm_fini(RdmaDeviceResources *dev_res, RdmaBackendDev *backend_dev,
res_tbl_free(&dev_res->cq_tbl);
res_tbl_free(&dev_res->pd_tbl);
if (dev_res->qp_hash) {
g_hash_table_destroy(dev_res->qp_hash);
}
}

View File

@ -43,7 +43,6 @@ static Property pvrdma_dev_properties[] = {
DEFINE_PROP_UINT64("dev-caps-max-mr-size", PVRDMADev, dev_attr.max_mr_size,
MAX_MR_SIZE),
DEFINE_PROP_INT32("dev-caps-max-qp", PVRDMADev, dev_attr.max_qp, MAX_QP),
DEFINE_PROP_INT32("dev-caps-max-sge", PVRDMADev, dev_attr.max_sge, MAX_SGE),
DEFINE_PROP_INT32("dev-caps-max-cq", PVRDMADev, dev_attr.max_cq, MAX_CQ),
DEFINE_PROP_INT32("dev-caps-max-mr", PVRDMADev, dev_attr.max_mr, MAX_MR),
DEFINE_PROP_INT32("dev-caps-max-pd", PVRDMADev, dev_attr.max_pd, MAX_PD),
@ -549,8 +548,9 @@ static void init_dev_caps(PVRDMADev *dev)
sizeof(struct pvrdma_rq_wqe_hdr));
dev->dev_attr.max_qp_wr = pg_tbl_bytes /
(wr_sz + sizeof(struct pvrdma_sge) * MAX_SGE) -
TARGET_PAGE_SIZE; /* First page is ring state */
(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) -
@ -626,8 +626,6 @@ static void pvrdma_realize(PCIDevice *pdev, Error **errp)
init_regs(pdev);
init_dev_caps(dev);
rc = init_msix(pdev, errp);
if (rc) {
goto out;
@ -640,6 +638,8 @@ static void pvrdma_realize(PCIDevice *pdev, Error **errp)
goto out;
}
init_dev_caps(dev);
rc = rdma_rm_init(&dev->rdma_dev_res, &dev->dev_attr, errp);
if (rc) {
goto out;

View File

@ -121,6 +121,16 @@ static void pvrdma_qp_ops_comp_handler(void *ctx, struct ibv_wc *wc)
g_free(ctx);
}
static void complete_with_error(uint32_t vendor_err, void *ctx)
{
struct ibv_wc wc = {0};
wc.status = IBV_WC_GENERAL_ERR;
wc.vendor_err = vendor_err;
pvrdma_qp_ops_comp_handler(ctx, &wc);
}
void pvrdma_qp_ops_fini(void)
{
rdma_backend_unregister_comp_handler();
@ -133,7 +143,7 @@ int pvrdma_qp_ops_init(void)
return 0;
}
int pvrdma_qp_send(PVRDMADev *dev, uint32_t qp_handle)
void pvrdma_qp_send(PVRDMADev *dev, uint32_t qp_handle)
{
RdmaRmQP *qp;
PvrdmaSqWqe *wqe;
@ -145,7 +155,8 @@ int pvrdma_qp_send(PVRDMADev *dev, uint32_t qp_handle)
qp = rdma_rm_get_qp(&dev->rdma_dev_res, qp_handle);
if (unlikely(!qp)) {
return -EINVAL;
pr_dbg("Invalid qpn\n");
return;
}
ring = (PvrdmaRing *)qp->opaque;
@ -168,7 +179,8 @@ int 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);
return -EIO;
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);
@ -179,7 +191,15 @@ int pvrdma_qp_send(PVRDMADev *dev, uint32_t qp_handle)
if (sgid_idx <= 0) {
pr_dbg("Fail to get bk sgid_idx for sgid_idx %d\n",
wqe->hdr.wr.ud.av.gid_index);
return -EIO;
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);
complete_with_error(VENDOR_ERR_INV_NUM_SGE, comp_ctx);
continue;
}
rdma_backend_post_send(&dev->backend_dev, &qp->backend_qp, qp->qp_type,
@ -193,11 +213,9 @@ int pvrdma_qp_send(PVRDMADev *dev, uint32_t qp_handle)
wqe = pvrdma_ring_next_elem_read(ring);
}
return 0;
}
int pvrdma_qp_recv(PVRDMADev *dev, uint32_t qp_handle)
void pvrdma_qp_recv(PVRDMADev *dev, uint32_t qp_handle)
{
RdmaRmQP *qp;
PvrdmaRqWqe *wqe;
@ -207,7 +225,8 @@ int pvrdma_qp_recv(PVRDMADev *dev, uint32_t qp_handle)
qp = rdma_rm_get_qp(&dev->rdma_dev_res, qp_handle);
if (unlikely(!qp)) {
return -EINVAL;
pr_dbg("Invalid qpn\n");
return;
}
ring = &((PvrdmaRing *)qp->opaque)[1];
@ -227,6 +246,13 @@ int pvrdma_qp_recv(PVRDMADev *dev, uint32_t qp_handle)
comp_ctx->cqe.qp = 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);
complete_with_error(VENDOR_ERR_INV_NUM_SGE, comp_ctx);
continue;
}
rdma_backend_post_recv(&dev->backend_dev, &dev->rdma_dev_res,
&qp->backend_qp, qp->qp_type,
(struct ibv_sge *)&wqe->sge[0], wqe->hdr.num_sge,
@ -236,8 +262,6 @@ int pvrdma_qp_recv(PVRDMADev *dev, uint32_t qp_handle)
wqe = pvrdma_ring_next_elem_read(ring);
}
return 0;
}
void pvrdma_cq_poll(RdmaDeviceResources *dev_res, uint32_t cq_handle)

View File

@ -20,8 +20,8 @@
int pvrdma_qp_ops_init(void);
void pvrdma_qp_ops_fini(void);
int pvrdma_qp_send(PVRDMADev *dev, uint32_t qp_handle);
int pvrdma_qp_recv(PVRDMADev *dev, uint32_t qp_handle);
void pvrdma_qp_send(PVRDMADev *dev, uint32_t qp_handle);
void pvrdma_qp_recv(PVRDMADev *dev, uint32_t qp_handle);
void pvrdma_cq_poll(RdmaDeviceResources *dev_res, uint32_t cq_handle);
#endif