Merge branch 'master' of https://github.com/daos-stack/daos into grom…

…72/pmdk-2.0.0

Required-githooks: true

.github/workflows/rpm-build-and-test.yml

@@ -46,10 +46,8 @@ jobs:
       dequoted_message: ${{ steps.dequoted_commit_message.outputs.text }}
     steps:
       - name: Checkout code
-        uses: actions/checkout@v3
+        uses: actions/checkout@v4
         with:
-          submodules: false
-          fetch-depth: 1
           ref: ${{ github.event.pull_request.head.sha }}
       - name: Import Commit Message
         id: commit_message
@@ -229,10 +227,8 @@ jobs:
             echo "STAGE_NAME=Build RPM on $DISTRO_NAME $DISTRO_VERSION" >> $GITHUB_ENV
             echo "FVERSION=$FVERSION" >> $GITHUB_ENV
       - name: Checkout code
-        uses: actions/checkout@v3
+        uses: actions/checkout@v4
         with:
-          submodules: true
-          fetch-depth: 500
           ref: ${{ github.event.pull_request.head.sha }}
       - name: Build RPM Docker image
         id: build-rpm-docker-image
@@ -315,35 +311,49 @@ jobs:
     steps:
       - name: Import commit pragmas
         uses: ./.github/actions/import-commit-pragmas
+      - name: Checkout code
+        uses: actions/checkout@v4
+        with:
+          ref: ${{ github.event.pull_request.head.sha }}
       - name: Calculate Functional Testing Matrix
         id: matrix
         run: |  # do not use the non-| format for this script
-             l=()
-             trap 'echo "text=[$(IFS=","; echo "${l[*]}")]" >> $GITHUB_OUTPUT' EXIT
-             if ${CP_SKIP_FUNC_TEST:-false}; then
-                 exit 0
-             fi
+             . ci/gha_functions.sh
+             set -eu
              # it might seem tempting to factor in the result of the build for this
-             # distro here and not include a failed build in the test matrix
+             # distro here and not include a failed build in the test matrix but
              # the problem with that is that if/when the user asks GHA to rebuild
              # all failed jobs and a previously failed RPM job is successful, the
              # test matrix won't include testing it since it was calculated and was
              # successful on the previous run without the failed build stage in it
-             if ! ${CP_SKIP_BUILD_EL8_RPM:-false} &&
-                ! ${CP_SKIP_FUNC_TEST_EL8:-false}; then
-                 # it would definitely be nicer to get these into the environment
-                 # as unquoted strings so that we didn't have to double quote here
-                 l+=('"el8"')
+             l=()
+             trap 'echo "text=[$(IFS=","; echo "${l[*]}")]" >> $GITHUB_OUTPUT' EXIT
+             if ${CP_SKIP_FUNC_TEST:-false}; then
+                 exit 0
              fi
-             if ! ${CP_SKIP_BUILD_EL9_RPM:-false} &&
-                ! ${CP_SKIP_FUNC_TEST_EL9:-false}; then
-                 l+=('"el9"')
+             if ! cd src/tests/ftest; then
+                 echo "src/tests/ftest doesn't exist."
+                 echo "Could not determine if tests exist for this stage, assuming they do."
+                 exit 0
              fi
-             if ${{ github.event_name == 'push' }} ||
-                (${{ github.event_name == 'pull_request' }} &&
-                ! ${CP_SKIP_BUILD_LEAP15_RPM:-false} &&
-                ! ${CP_SKIP_FUNC_TEST_LEAP15:-true}); then
-                 l+=('"leap15"')
+
+             if ./launch.py --list "$(get_test_tags "-hw")"; then
+                 if ! ${CP_SKIP_BUILD_EL8_RPM:-false} &&
+                    ! ${CP_SKIP_FUNC_TEST_EL8:-false}; then
+                     # it would definitely be nicer to get these into the environment
+                     # as unquoted strings so that we didn't have to double quote here
+                     l+=('"el8"')
+                 fi
+                 if ! ${CP_SKIP_BUILD_EL9_RPM:-false} &&
+                    ! ${CP_SKIP_FUNC_TEST_EL9:-false}; then
+                     l+=('"el9"')
+                 fi
+                 if ${{ github.event_name == 'push' }} ||
+                    (${{ github.event_name == 'pull_request' }} &&
+                    ! ${CP_SKIP_BUILD_LEAP15_RPM:-false} &&
+                    ! ${CP_SKIP_FUNC_TEST_LEAP15:-true}); then
+                     l+=('"leap15"')
+                 fi
              fi
 
   Functional:
@@ -400,6 +410,7 @@ jobs:
             env
             STAGE_TAGS="-hw"
             FTEST_ARG=""
+            INST_RPMS="daos-client daos-tests daos-server daos-serialize daos-tests-internal"
             case "${{ matrix.distro }}" in
                 'el8')
                     CHROOT_NAME="rocky+epel-8-x86_64"
@@ -456,10 +467,9 @@ jobs:
             echo "${PROV_DISTRO_VERSION_MAJOR:-$DISTRO_VERSION_MAJOR}_LOCAL_REPO=not_used" >> \
                  $GITHUB_ENV
             echo "LABEL=$LABEL" >> $GITHUB_ENV
-            echo -n "INST_RPMS=daos-client daos-tests daos-server daos-serialize" >> $GITHUB_ENV
-            echo -n " daos-tests-internal" >> $GITHUB_ENV
+            echo "INST_RPMS=$INST_RPMS" >> $GITHUB_ENV
       - name: Checkout code
-        uses: actions/checkout@v3
+        uses: actions/checkout@v4
         with:
           submodules: true
           fetch-depth: 500
@@ -532,34 +542,48 @@ jobs:
     steps:
       - name: Import commit pragmas
         uses: ./.github/actions/import-commit-pragmas
+      - name: Checkout code
+        uses: actions/checkout@v4
+        with:
+          ref: ${{ github.event.pull_request.head.sha }}
       - name: Calculate Functional Testing Matrix
         id: matrix
         run: |  # do not use the non-| format for this script
+             . ci/gha_functions.sh
              set -eu
-             l=()
-             trap 'echo "text=[$(IFS=","; echo "${l[*]}")]" >> $GITHUB_OUTPUT' EXIT
-             if ${CP_SKIP_FUNC_HW_TEST:-false}; then
-                 exit 0
-             fi
              # it might seem tempting to factor in the result of the build for this
-             # distro here and not include a failed build in the test matrix
+             # distro here and not include a failed build in the test matrix but
              # the problem with that is that if/when the user asks GHA to rebuild
              # all faiiled jobs and a previously failed RPM job is successful, the
              # test matrix won't include testing it since it was calculated and was
              # successful on the previous run without the failed build stage in it
-             if ! "${CP_SKIP_FUNC_HW_TEST_LARGE:-false}"; then
+             l=()
+             trap 'echo "text=[$(IFS=","; echo "${l[*]}")]" >> $GITHUB_OUTPUT' EXIT
+             if ${CP_SKIP_FUNC_HW_TEST:-false}; then
+                 exit 0
+             fi
+             if ! cd src/tests/ftest; then
+                 echo "src/tests/ftest doesn't exist."
+                 echo "Could not determine if tests exist for this stage, assuming they do."
+                 exit 0
+             fi
+             if ! "${CP_SKIP_FUNC_HW_TEST_LARGE:-false}" &&
+                ./launch.py --list "$(get_test_tags "hw,large,-provider")"; then
                  # it would definitely be nicer to get these into the environment
                  # as unquoted strings so that we didn't have to double quote here
                  l+=('"Large"')
              fi
-             if ! ${CP_SKIP_FUNC_HW_TEST_MEDIUM:-false}; then
+             if ! ${CP_SKIP_FUNC_HW_TEST_MEDIUM:-false} &&
+                ./launch.py --list "$(get_test_tags "hw,medium,-provider")"; then
                  l+=('"Medium"')
              fi
-             if ! ${CP_SKIP_FUNC_HW_TEST_MEDIUM_VERBS_PROVIDER:-false}; then
+             if ! ${CP_SKIP_FUNC_HW_TEST_MEDIUM_VERBS_PROVIDER:-false} &&
+                ./launch.py --list "$(get_test_tags "hw,medium,provider")"; then
                  l+=('"Medium Verbs Provider"')
              fi
              if ${{ github.event_name == 'push' }} &&
-                ! ${CP_SKIP_FUNC_HW_TEST_MEDIUM_UCX_PROVIDER:-false}; then
+                ! ${CP_SKIP_FUNC_HW_TEST_MEDIUM_UCX_PROVIDER:-false} &&
+                ./launch.py --list "$(get_test_tags "hw,medium,provider")"; then
                  l+=('"Medium UCX Provider"')
              fi
 
@@ -581,7 +605,7 @@ jobs:
       (needs.Build-RPM.result == 'success' ||
        needs.Build-RPM.result == 'skipped') &&
       (needs.Functional.result == 'success' ||
-      needs.Functional.result == 'skipped')
+       needs.Functional.result == 'skipped')
     env:
       CONFIG_POWER_ONLY: false
       PRAGMA_SUFFIX: -vm
@@ -614,6 +638,7 @@ jobs:
         run: |
             STAGE_TAGS="hw"
             FTEST_ARG="--nvme=auto:-3DNAND"
+            INST_RPMS="daos-client daos-tests daos-server daos-serialize daos-tests-internal"
             CHROOT_NAME="rocky+epel-8-x86_64"
             DISTRO_NAME="EL"
             DISTRO_NAME_UPPER="EL"
@@ -634,8 +659,19 @@ jobs:
                 STAGE_TAGS+=",large"
                 SIZE="LARGE"
             fi
-            if [[ "${{ matrix.stage }}" = *\ Prvider ]]; then
+            if [[ "${{ matrix.stage }}" = *\ Provider ]]; then
                 STAGE_TAGS+=",provider"
+                if [[ "${{ matrix.stage }}" = *\ Verbs\ * ]]; then
+                  FTEST_ARG+=' --provider ofi+verbs'
+                elif [[ "${{ matrix.stage }}" = *\ UCX\ * ]]; then
+                  FTEST_ARG+=' --provider ucx+dc_x'
+                  INST_RPMS+=' mercury-ucx'
+                elif [[ "${{ matrix.stage }}" = *\ TCP\ * ]]; then
+                  FTEST_ARG+=' --provider ofi+tcp'
+                else
+                    echo "Unknown provider in ${{ matrix.stage }}"
+                    exit 1
+                fi
             else
                 STAGE_TAGS+=",-provider"
             fi
@@ -652,11 +688,10 @@ jobs:
             echo "${PROV_DISTRO_VERSION_MAJOR:-$DISTRO_VERSION_MAJOR}_LOCAL_REPO=not_used" >> \
                  $GITHUB_ENV
             echo "LABEL=$LABEL" >> $GITHUB_ENV
-            echo -n "INST_RPMS=daos-client daos-tests daos-server daos-serialize" >> $GITHUB_ENV
-            echo " daos-tests-internal" >> $GITHUB_ENV
+            echo "INST_RPMS=$INST_RPMS" >> $GITHUB_ENV
             echo "SIZE=$SIZE" >> $GITHUB_ENV
       - name: Checkout code
-        uses: actions/checkout@v3
+        uses: actions/checkout@v4
         with:
           submodules: true
           fetch-depth: 500

src/cart/crt_rpc.h

@@ -216,11 +216,11 @@ crt_rpc_unlock(struct crt_rpc_priv *rpc_priv)
 	D_MUTEX_UNLOCK(&rpc_priv->crp_mutex);
 }
 
-#define CRT_PROTO_INTERNAL_VERSION 5
+#define CRT_PROTO_INTERNAL_VERSION 4
 #define CRT_PROTO_FI_VERSION 3
 #define CRT_PROTO_ST_VERSION 1
 #define CRT_PROTO_CTL_VERSION 1
-#define CRT_PROTO_IV_VERSION 1
+#define CRT_PROTO_IV_VERSION       2
 
 /* LIST of internal RPCS in form of:
  * OPCODE, flags, FMT, handler, corpc_hdlr,

src/client/dfuse/dfuse.h

@@ -689,7 +689,7 @@ struct fuse_lowlevel_ops dfuse_ops;
 	do {                                                                                       \
 		int    __rc;                                                                       \
 		double timeout = 0;                                                                \
-		if (atomic_load_relaxed(&(ie)->ie_open_count) == 0) {                              \
+		if (atomic_load_relaxed(&(ie)->ie_il_count) == 0) {                                \
 			timeout = (ie)->ie_dfs->dfc_attr_timeout;                                  \
 			dfuse_mcache_set_time(ie);                                                 \
 		}                                                                                  \

src/client/dfuse/dfuse_fuseops.c

@@ -372,6 +372,11 @@ df_ll_symlink(fuse_req_t req, const char *link, fuse_ino_t parent, const char *n
 	DFUSE_REPLY_ERR_RAW(dfuse_info, req, rc);
 }
 
+/* Do not allow security xattrs to be set or read, see DAOS-14639 */
+#define XATTR_SEC   "security."
+/* Do not allow either system.posix_acl_default or system.posix_acl_access */
+#define XATTR_P_ACL "system.posix_acl"
+
 void
 df_ll_setxattr(fuse_req_t req, fuse_ino_t ino, const char *name, const char *value, size_t size,
 	       int flags)
@@ -385,6 +390,12 @@ df_ll_setxattr(fuse_req_t req, fuse_ino_t ino, const char *name, const char *val
 		D_GOTO(err, rc = EPERM);
 	}
 
+	if (strncmp(name, XATTR_SEC, sizeof(XATTR_SEC) - 1) == 0)
+		D_GOTO(err, rc = ENOTSUP);
+
+	if (strncmp(name, XATTR_P_ACL, sizeof(XATTR_P_ACL) - 1) == 0)
+		D_GOTO(err, rc = ENOTSUP);
+
 	inode = dfuse_inode_lookup_nf(dfuse_info, ino);
 
 	if (!inode->ie_dfs->dfs_ops->setxattr)
@@ -406,6 +417,12 @@ df_ll_getxattr(fuse_req_t req, fuse_ino_t ino, const char *name, size_t size)
 	struct dfuse_inode_entry *inode;
 	int                       rc;
 
+	if (strncmp(name, XATTR_SEC, sizeof(XATTR_SEC) - 1) == 0)
+		D_GOTO(err, rc = ENODATA);
+
+	if (strncmp(name, XATTR_P_ACL, sizeof(XATTR_P_ACL) - 1) == 0)
+		D_GOTO(err, rc = ENODATA);
+
 	inode = dfuse_inode_lookup_nf(dfuse_info, ino);
 
 	if (!inode->ie_dfs->dfs_ops->getxattr)

src/client/pydaos/raw/conversion.py

@@ -8,6 +8,8 @@
 import ctypes
 import uuid
 
+from .. import DAOS_MAGIC, pydaos_shim
+
 
 def c_uuid_to_str(cuuid):
     """ utility function to convert a C uuid into a standard string format """
@@ -29,3 +31,19 @@ def str_to_c_uuid(uuidstr):
     cuuid = (ctypes.c_ubyte * 16)()
     c_uuid(puuid, cuuid)
     return cuuid
+
+
+def c_err_to_str(err_num):
+    """ Utility function to convert C error code to its string id.
+
+    Args:
+        err_num (int): C error code.
+
+    Returns:
+        str: String identifying a C error code.
+
+    """
+    err_str = pydaos_shim.err_to_str(DAOS_MAGIC, err_num)
+    if err_str:
+        return err_str
+    return str(err_num)

src/container/srv_target.c

@@ -835,8 +835,6 @@ ds_cont_child_stop_all(struct ds_pool_child *pool_child)
 		DP_UUID(pool_child->spc_uuid),
 		dss_get_module_info()->dmi_tgt_id);
 
-	D_ASSERT(d_list_empty(&pool_child->spc_list));
-
 	cont_list = &pool_child->spc_cont_list;
 	while (!d_list_empty(cont_list)) {
 		cont_child = d_list_entry(cont_list->next,
@@ -1532,7 +1530,7 @@ ds_cont_local_open(uuid_t pool_uuid, uuid_t cont_hdl_uuid, uuid_t cont_uuid,
 		if (ddra == NULL)
 			D_GOTO(err_dtx, rc = -DER_NOMEM);
 
-		ddra->pool = ds_pool_child_get(hdl->sch_cont->sc_pool);
+		ddra->pool = ds_pool_child_lookup(hdl->sch_cont->sc_pool->spc_uuid);
 		uuid_copy(ddra->co_uuid, cont_uuid);
 		rc = dss_ult_create(ds_dtx_resync, ddra, DSS_XS_SELF,
 				    0, 0, NULL);

src/include/daos_srv/pool.h

@@ -58,6 +58,7 @@ struct ds_pool {
 	ABT_cond		sp_fetch_hdls_cond;
 	ABT_cond		sp_fetch_hdls_done_cond;
 	struct ds_iv_ns		*sp_iv_ns;
+	uint32_t		*sp_states;	/* pool child state array */
 
 	/* structure related to EC aggregate epoch query */
 	d_list_t		sp_ec_ephs_list;
@@ -128,6 +129,13 @@ struct ds_pool_hdl {
 struct ds_pool_hdl *ds_pool_hdl_lookup(const uuid_t uuid);
 void ds_pool_hdl_put(struct ds_pool_hdl *hdl);
 
+enum pool_child_state {
+	POOL_CHILD_NEW	= 0,
+	POOL_CHILD_STARTING,
+	POOL_CHILD_STARTED,
+	POOL_CHILD_STOPPING,
+};
+
 /*
  * Per-thread pool object
  *
@@ -160,8 +168,9 @@ struct ds_pool_child {
 	int		spc_ref;
 	ABT_eventual	spc_ref_eventual;
 
-	uint64_t	spc_discard_done:1;
+	uint32_t	spc_discard_done:1;
 	uint32_t	spc_reint_mode;
+	uint32_t	*spc_state;	/* Pointer to ds_pool->sp_states[i] */
 	/**
 	 * Per-pool per-module metrics, see ${modname}_pool_metrics for the
 	 * actual structure. Initialized only for modules that specified a
@@ -182,9 +191,16 @@ struct ds_pool_svc_op_val {
 	char ov_resvd[60];
 };
 
+/* Find ds_pool_child in cache, hold one reference */
 struct ds_pool_child *ds_pool_child_lookup(const uuid_t uuid);
-struct ds_pool_child *ds_pool_child_get(struct ds_pool_child *child);
+/* Put the reference held by ds_pool_child_lookup() */
 void ds_pool_child_put(struct ds_pool_child *child);
+/* Start ds_pool child */
+int ds_pool_child_start(uuid_t pool_uuid);
+/* Stop ds_pool_child */
+int ds_pool_child_stop(uuid_t pool_uuid);
+/* Query pool child state */
+uint32_t ds_pool_child_state(struct ds_pool *pool, uint32_t tgt_id);
 
 int ds_pool_bcast_create(crt_context_t ctx, struct ds_pool *pool,
 			 enum daos_module_id module, crt_opcode_t opcode,

src/pool/srv_pool.c

@@ -2755,7 +2755,7 @@ ds_pool_create_handler(crt_rpc_t *rpc)
 
 	rc = pool_prop_default_copy(prop_dup, prop);
 	if (rc) {
-		DL_ERROR(rc, "daos_prop_default_copy failed.\n");
+		DL_ERROR(rc, "daos_prop_default_copy() failed");
 		D_GOTO(out_tx, rc);
 	}