Skip to contentSkip to conversation

checking for owned ports in portmask

PATCH Superseded Stephen Hemminger 4 participants 31 messages Mar 12, 2020 – Apr 10, 2020

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

cover checking for owned ports in portmask

The failsafe, bonding, and netvsc PMD use eth dev port ownership to control sub devices. These sub devices are hidden in the normal iteration over ports but some applications use a direct port mask. In these cases, user may still (incorrectly) try to directly use the sub devices.

This patch set introduces a helper function to check if a port is under ownership. And the follow on patches use that.

IMHO this should not be marked as experimental and done as a bug fix back to DPDK 18.11 because then it will get picked up in LTS and users won't run into it on Azure.

An alternative proposed in earlier discussion is to have each application try and own the port. But this solution would be more invasive, and does not handle the case if secondary process exits prematurely without releasing ownership. It makes more sense to keep the concept of ownership as strictly part of the device model and not part of the application layer.

Stephen Hemminger (7):
ethdev: add function to test port ownership examples/l2fwd-cat: block attempts to use owned ports examples/l3fwd: block attempts to use owned ports examples/l3fwd-acl: block attempts to use owned ports examples/l3fwd-power: block attempts to use owned ports examples/tep_termination: block attempts to use owned ports examples/vhost: block attempts to use owned ports

 examples/l2fwd-cat/Makefile              |  2 ++
 examples/l2fwd-cat/l2fwd-cat.c           |  3 +++
 examples/l2fwd-cat/meson.build           |  1 +
 examples/l3fwd-acl/Makefile              |  3 +++
 examples/l3fwd-acl/main.c                |  4 ++++
 examples/l3fwd-acl/meson.build           |  1 +
 examples/l3fwd-power/main.c              |  4 ++++
 examples/l3fwd/Makefile                  |  3 +++
 examples/l3fwd/main.c                    |  4 ++++
 examples/l3fwd/meson.build               |  1 +
 examples/tep_termination/Makefile        |  2 ++
 examples/tep_termination/main.c          |  2 +-
 examples/tep_termination/meson.build     |  1 +
 examples/vhost/main.c                    |  2 +-
 lib/librte_ethdev/rte_ethdev.c           |  9 +++++++++
 lib/librte_ethdev/rte_ethdev.h           | 15 +++++++++++++++
 lib/librte_ethdev/rte_ethdev_version.map |  3 +++
 17 files changed, 58 insertions(+), 2 deletions(-)
Signature · 1 line
2.20.1
#1 lore raw

PATCH 1/7 ethdev: add function to test port ownership

Applications using a port mask need a method to be able to test for (and reject) ethdev ports that are in use for other purposes.

Fixes 5b7ba31148a8 ("ethdev: add port ownership")
---
 lib/librte_ethdev/rte_ethdev.c           |  9 +++++++++
 lib/librte_ethdev/rte_ethdev.h           | 15 +++++++++++++++
 lib/librte_ethdev/rte_ethdev_version.map |  3 +++
 3 files changed, 27 insertions(+)
lib/librte_ethdev/rte_ethdev.c +9 -0
index 774c721b3484..38f99f417dba 100644
@@ -685,6 +685,15 @@ rte_eth_dev_owner_unset(const uint16_t port_id, const uint64_t owner_id)
 	return ret;
 }
 
+int
+rte_eth_dev_is_owned_by(uint16_t port_id, uint64_t owner_id)
+{
+	if (!rte_eth_dev_is_valid_port(port_id))
+		return 0;
+
+	return rte_eth_devices[port_id].data->owner.id == owner_id;
+}
+
 int
 rte_eth_dev_owner_delete(const uint64_t owner_id)
 {
lib/librte_ethdev/rte_ethdev.h +15 -0
index d1a593ad112a..91a8f9578282 100644
@@ -1530,6 +1530,21 @@ uint64_t rte_eth_find_next_owned_by(uint16_t port_id,
 	     (unsigned int)p < (unsigned int)RTE_MAX_ETHPORTS; \
 	     p = rte_eth_find_next_owned_by(p + 1, o))
 
+/**
+ * Test if a port is owned
+ *
+ * @param port_id
+ *   The port identifier of the Ethernet device
+ * @param	owner_id
+ *   The owner identifier.
+ *   RTE_ETH_DEV_NO_OWNER means test if port is not owned.
+ * @return
+ *   - 0 if port is out of range or not owned by owner_id
+ *   - 1 if device is associated with owner_id
+ */
+__rte_experimental
+int rte_eth_dev_is_owned_by(uint16_t port_id, uint64_t owner_id);
+
 /**
  * Iterates over valid ethdev ports.
  *
lib/librte_ethdev/rte_ethdev_version.map +3 -0
index 3f32fdecf722..95231fe3410c 100644
@@ -230,4 +230,7 @@ EXPERIMENTAL {
 
 	# added in 20.02
 	rte_flow_dev_dump;
+
+	# added in 20.05
+	rte_eth_dev_is_owned_by;
 };
Signature · 1 line
2.20.1
#2 lore raw

Re: [dpdk-dev] [PATCH 1/7] ethdev: add function to test port ownership

Hi Stephen

From: Stephen Hemminger

Quoted 3 lines

Applications using a port mask need a method to be able to test for (and reject) ethdev ports that are in use for other purposes.

There is already function which is thread safe to get this information:
rte_eth_dev_owner_get.

I don't think we need one more.

Quoted 70 lines
Fixes 5b7ba31148a8 ("ethdev: add port ownership")
---
 lib/librte_ethdev/rte_ethdev.c           |  9 +++++++++
 lib/librte_ethdev/rte_ethdev.h           | 15 +++++++++++++++
 lib/librte_ethdev/rte_ethdev_version.map |  3 +++
 3 files changed, 27 insertions(+)
lib/librte_ethdev/rte_ethdev.c +7 -0
index 774c721b3484..38f99f417dba 100644
@@ -685,6 +685,15 @@ rte_eth_dev_owner_unset(const uint16_t port_id,
 const uint64_t owner_id)
 	return ret;
 }
 
+int
+rte_eth_dev_is_owned_by(uint16_t port_id, uint64_t owner_id) {
+	if (!rte_eth_dev_is_valid_port(port_id))
+		return 0;
+
+	return rte_eth_devices[port_id].data->owner.id == owner_id; }
+
 int
 rte_eth_dev_owner_delete(const uint64_t owner_id)  { diff --git

a/lib/librte_ethdev/rte_ethdev.h b/lib/librte_ethdev/rte_ethdev.h index d1a593ad112a..91a8f9578282 100644

lib/librte_ethdev/rte_ethdev.h +15 -0
@@ -1530,6 +1530,21 @@ uint64_t rte_eth_find_next_owned_by(uint16_t
 port_id,
 	     (unsigned int)p < (unsigned int)RTE_MAX_ETHPORTS; \
 	     p = rte_eth_find_next_owned_by(p + 1, o))
 
+/**
+ * Test if a port is owned
+ *
+ * @param port_id
+ *   The port identifier of the Ethernet device
+ * @param	owner_id
+ *   The owner identifier.
+ *   RTE_ETH_DEV_NO_OWNER means test if port is not owned.
+ * @return
+ *   - 0 if port is out of range or not owned by owner_id
+ *   - 1 if device is associated with owner_id
+ */
+__rte_experimental
+int rte_eth_dev_is_owned_by(uint16_t port_id, uint64_t owner_id);
+
 /**
  * Iterates over valid ethdev ports.

*

lib/librte_ethdev/rte_ethdev_version.map +3 -1
index 3f32fdecf722..95231fe3410c 100644
@@ -230,4 +230,7 @@ EXPERIMENTAL {
 
 	# added in 20.02
 	rte_flow_dev_dump;
+
+	# added in 20.05
+	rte_eth_dev_is_owned_by;
 };
--

2.20.1

#3 lore raw

PATCH 2/7 examples/l2fwd-cat: block attempts to use owned ports

In reply to Stephen Hemminger

If a ethdev port is in use for a sub device, then it should not be allowed in the portmask of application.

Fixes 5b7ba31148a8 ("ethdev: add port ownership")
---
 examples/l2fwd-cat/Makefile    | 2 ++
 examples/l2fwd-cat/l2fwd-cat.c | 3 +++
 examples/l2fwd-cat/meson.build | 1 +
 3 files changed, 6 insertions(+)
examples/l2fwd-cat/Makefile +2 -0
index b0e53c37e8d8..aa19347c1545 100644
@@ -20,6 +20,7 @@ static: build/$(APP)-static
 PKGCONF ?= pkg-config
 
 PC_FILE := $(shell $(PKGCONF) --path libdpdk 2>/dev/null)
+CFLAGS += -DALLOW_EXPERIMENTAL_API
 CFLAGS += -O3 $(shell $(PKGCONF) --cflags libdpdk)
 LDFLAGS_SHARED = $(shell $(PKGCONF) --libs libdpdk)
 LDFLAGS_STATIC = -Wl,-Bstatic $(shell $(PKGCONF) --static --libs libdpdk)
@@ -65,6 +66,7 @@ endif
 
 EXTRA_CFLAGS += -O3 -g -Wfatal-errors
 
+CFLAGS += -DALLOW_EXPERIMENTAL_API
 CFLAGS += -I$(PQOS_INSTALL_PATH)/../include
 
 LDLIBS += -L$(PQOS_INSTALL_PATH)
examples/l2fwd-cat/l2fwd-cat.c +3 -0
index 6838f288c621..c1c59b2d81f2 100644
@@ -42,6 +42,9 @@ port_init(uint16_t port, struct rte_mempool *mbuf_pool)
 	if (!rte_eth_dev_is_valid_port(port))
 		return -1;
 
+	if (!rte_eth_dev_is_owned_by(portid, RTE_ETH_DEV_NO_OWNER)) {
+		return -1;
+
 	/* Configure the Ethernet device. */
 	retval = rte_eth_dev_configure(port, rx_rings, tx_rings, &port_conf);
 	if (retval != 0)
examples/l2fwd-cat/meson.build +1 -0
index 4e2777a03358..5c062c1354e3 100644
@@ -6,6 +6,7 @@
 # To build this example as a standalone application with an already-installed
 # DPDK instance, use 'make'
 
+allow_experimental_apis = true
 pqos = cc.find_library('pqos', required: false)
 build = pqos.found()
 ext_deps += pqos
Signature · 1 line
2.20.1
#4 lore raw

Re: [dpdk-dev] [PATCH 2/7] examples/l2fwd-cat: block attempts to use owned ports

Hi Stephen

From: Stephen Hemminger

I think the correct terminology here is (can be the title):
Make the application be port ownership aware.

Some applications may be ownership aware and others not.

Quoted 60 lines

If a ethdev port is in use for a sub device, then it should not be allowed in the portmask of application.

Fixes 5b7ba31148a8 ("ethdev: add port ownership")
---
 examples/l2fwd-cat/Makefile    | 2 ++
 examples/l2fwd-cat/l2fwd-cat.c | 3 +++
 examples/l2fwd-cat/meson.build | 1 +
 3 files changed, 6 insertions(+)
examples/l2fwd-cat/Makefile +1 -0
index b0e53c37e8d8..aa19347c1545 100644
@@ -20,6 +20,7 @@ static: build/$(APP)-static  PKGCONF ?= pkg-config
 
 PC_FILE := $(shell $(PKGCONF) --path libdpdk 2>/dev/null)
+CFLAGS += -DALLOW_EXPERIMENTAL_API
 CFLAGS += -O3 $(shell $(PKGCONF) --cflags libdpdk)  LDFLAGS_SHARED =

$(shell $(PKGCONF) --libs libdpdk) LDFLAGS_STATIC = -Wl,-Bstatic $(shell $(PKGCONF) --static --libs libdpdk) @@ -65,6 +66,7 @@ endif

EXTRA_CFLAGS += -O3 -g -Wfatal-errors

examples/l2fwd-cat/Makefile +1 -0
+CFLAGS += -DALLOW_EXPERIMENTAL_API
 CFLAGS += -I$(PQOS_INSTALL_PATH)/../include
 
 LDLIBS += -L$(PQOS_INSTALL_PATH)
examples/l2fwd-cat/l2fwd- renamed from examples/l2fwd-cat/l2fwd-cat.c +0 -0

cat.c index 6838f288c621..c1c59b2d81f2 100644

examples/l2fwd-cat/l2fwd-cat.c +3 -0
@@ -42,6 +42,9 @@ port_init(uint16_t port, struct rte_mempool
 *mbuf_pool)
 	if (!rte_eth_dev_is_valid_port(port))
 		return -1;
 
+	if (!rte_eth_dev_is_owned_by(portid, RTE_ETH_DEV_NO_OWNER))
 {
+		return -1;
+
 	/* Configure the Ethernet device. */

retval = rte_eth_dev_configure(port, rx_rings, tx_rings,
&port_conf);

examples/l2fwd-cat/l2fwd-cat.c +0 -0
 	if (retval != 0)
examples/l2fwd- renamed from examples/l2fwd-cat/meson.build +0 -0

cat/meson.build index 4e2777a03358..5c062c1354e3 100644

examples/l2fwd-cat/meson.build +2 -1
@@ -6,6 +6,7 @@
 # To build this example as a standalone application with an already-installed
 # DPDK instance, use 'make'
 
+allow_experimental_apis = true
 pqos = cc.find_library('pqos', required: false)  build = pqos.found()  ext_deps
+= pqos
--

2.20.1

#5 lore raw

PATCH 3/7 examples/l3fwd: block attempts to use owned ports

In reply to Stephen Hemminger

If a ethdev port is in use for a sub device, then it should not be allowed in the portmask of l3fwd.

Fixes 5b7ba31148a8 ("ethdev: add port ownership")
---
 examples/l3fwd/Makefile    | 3 +++
 examples/l3fwd/main.c      | 4 ++++
 examples/l3fwd/meson.build | 1 +
 3 files changed, 8 insertions(+)
examples/l3fwd/Makefile +3 -0
index 59a110d12a49..59cd00abe6e9 100644
@@ -25,6 +25,8 @@ CFLAGS += -O3 $(shell $(PKGCONF) --cflags libdpdk)
 LDFLAGS_SHARED = $(shell $(PKGCONF) --libs libdpdk)
 LDFLAGS_STATIC = -Wl,-Bstatic $(shell $(PKGCONF) --static --libs libdpdk)
 
+CFLAGS += -DALLOW_EXPERIMENTAL_API
+
 build/$(APP)-shared: $(SRCS-y) Makefile $(PC_FILE) | build
 	$(CC) $(CFLAGS) $(SRCS-y) -o $@ $(LDFLAGS) $(LDFLAGS_SHARED)
 
@@ -50,6 +52,7 @@ RTE_TARGET ?= $(notdir $(abspath $(dir $(firstword $(wildcard $(RTE_SDK)/*/.conf
 
 include $(RTE_SDK)/mk/rte.vars.mk
 
+CFLAGS += -DALLOW_EXPERIMENTAL_API
 CFLAGS += -I$(SRCDIR)
 CFLAGS += -O3 $(USER_FLAGS)
 CFLAGS += $(WERROR_FLAGS)
examples/l3fwd/main.c +4 -0
index dda430d68a20..307a490fb552 100644
@@ -221,6 +221,10 @@ check_port_config(void)
 			printf("port %u is not present on the board\n", portid);
 			return -1;
 		}
+		if (!rte_eth_dev_is_owned_by(portid, RTE_ETH_DEV_NO_OWNER)) {
+			printf("port %u is already in use\n", portid);
+			return -1;
+		}
 	}
 	return 0;
 }
examples/l3fwd/meson.build +1 -0
index ebed3b518c4b..7d72b1b365ae 100644
@@ -6,6 +6,7 @@
 # To build this example as a standalone application with an already-installed
 # DPDK instance, use 'make'
 
+allow_experimental_apis = true
 deps += ['hash', 'lpm', 'eventdev']
 sources = files(
 	'l3fwd_em.c', 'l3fwd_lpm.c', 'l3fwd_event.c',
Signature · 1 line
2.20.1
#6 lore raw

PATCH 4/7 examples/l3fwd-acl: block attempts to use owned ports

In reply to Stephen Hemminger

If a ethdev port is in use for a sub device, then it should not be allowed in the portmask of application.

Fixes 5b7ba31148a8 ("ethdev: add port ownership")
---
 examples/l3fwd-acl/Makefile    | 3 +++
 examples/l3fwd-acl/main.c      | 4 ++++
 examples/l3fwd-acl/meson.build | 1 +
 3 files changed, 8 insertions(+)
examples/l3fwd-acl/Makefile +3 -0
index d9909584b1c6..ff14b4fdd593 100644
@@ -20,7 +20,9 @@ static: build/$(APP)-static
 PKGCONF ?= pkg-config
 
 PC_FILE := $(shell $(PKGCONF) --path libdpdk 2>/dev/null)
+CFLAGS += -DALLOW_EXPERIMENTAL_API
 CFLAGS += -O3 $(shell $(PKGCONF) --cflags libdpdk)
+
 LDFLAGS_SHARED = $(shell $(PKGCONF) --libs libdpdk)
 LDFLAGS_STATIC = -Wl,-Bstatic $(shell $(PKGCONF) --static --libs libdpdk)
 
@@ -51,6 +53,7 @@ include $(RTE_SDK)/mk/rte.vars.mk
 
 CFLAGS += -O3
 CFLAGS += $(WERROR_FLAGS)
+CFLAGS += -DALLOW_EXPERIMENTAL_API
 
 # workaround for a gcc bug with noreturn attribute
 # http://gcc.gnu.org/bugzilla/show_bug.cgi?id=12603
examples/l3fwd-acl/main.c +4 -0
index fa92a2829740..695121bb7bb5 100644
@@ -1470,6 +1470,10 @@ check_port_config(void)
 			printf("port %u is not present on the board\n", portid);
 			return -1;
 		}
+		if (!rte_eth_dev_is_owned_by(portid, RTE_ETH_DEV_NO_OWNER)) {
+			printf("port %u is already in use\n", portid);
+			return -1;
+		}
 	}
 	return 0;
 }
examples/l3fwd-acl/meson.build +1 -0
index 7096e00c1073..6fa468b3aa9c 100644
@@ -6,6 +6,7 @@
 # To build this example as a standalone application with an already-installed
 # DPDK instance, use 'make'
 
+allow_experimental_apis = true
 deps += ['acl', 'lpm', 'hash']
 sources = files(
 	'main.c'
Signature · 1 line
2.20.1
#7 lore raw

PATCH 5/7 examples/l3fwd-power: block attempts to use owned ports

In reply to Stephen Hemminger

If a ethdev port is in use for a sub device, then it should not be allowed in the portmask of l3fwd.

Fixes 5b7ba31148a8 ("ethdev: add port ownership")
---
 examples/l3fwd-power/main.c | 4 ++++
 1 file changed, 4 insertions(+)
examples/l3fwd-power/main.c +4 -0
index c7fe0ec03495..3a8a7277cb0a 100644
@@ -1396,6 +1396,10 @@ check_port_config(void)
 								portid);
 			return -1;
 		}
+		if (!rte_eth_dev_is_owned_by(portid, RTE_ETH_DEV_NO_OWNER)) {
+			printf("port %u is already in use\n", portid);
+			return -1;
+		}
 	}
 	return 0;
 }
Signature · 1 line
2.20.1
#8 lore raw

PATCH 6/7 examples/tep_termination: block attempts to use owned ports

In reply to Stephen Hemminger

If a ethdev port is in use for a sub device, then it should not be allowed in the portmask of application.

Fixes 5b7ba31148a8 ("ethdev: add port ownership")
---
 examples/tep_termination/Makefile    | 2 ++
 examples/tep_termination/main.c      | 2 +-
 examples/tep_termination/meson.build | 1 +
 3 files changed, 4 insertions(+), 1 deletion(-)
examples/tep_termination/Makefile +2 -0
index 645112498d33..f55331fb24c2 100644
@@ -26,6 +26,7 @@ CFLAGS += -O3 $(shell $(PKGCONF) --cflags libdpdk)
 LDFLAGS_SHARED = $(shell $(PKGCONF) --libs libdpdk)
 LDFLAGS_STATIC = -Wl,-Bstatic $(shell $(PKGCONF) --static --libs libdpdk)
 
+CFLAGS += -DALLOW_EXPERIMENTAL_API
 CFLAGS += -Wno-deprecated-declarations
 
 build/$(APP)-shared: $(SRCS-y) Makefile $(PC_FILE) | build
@@ -61,6 +62,7 @@ endif
 CFLAGS += -O3
 CFLAGS += $(WERROR_FLAGS)
 CFLAGS += -Wno-deprecated-declarations
+CFLAGS += -DALLOW_EXPERIMENTAL_API
 
 include $(RTE_SDK)/mk/rte.extapp.mk
 endif
examples/tep_termination/main.c +1 -1
index ab956ad7ce96..928cd226d25a 100644
@@ -507,7 +507,7 @@ check_ports_num(unsigned max_nb_ports)
 	}
 
 	for (portid = 0; portid < nb_ports; portid++) {
-		if (!rte_eth_dev_is_valid_port(ports[portid])) {
+		if (!rte_eth_dev_is_owned_by(portid, RTE_ETH_DEV_NO_OWNER)) {
 			RTE_LOG(INFO, VHOST_PORT,
 				"\nSpecified port ID(%u) is not valid\n",
 				ports[portid]);
examples/tep_termination/meson.build +1 -0
index f65d689802a3..ddadf1c2aede 100644
@@ -9,6 +9,7 @@
 if not is_linux
 	build = false
 endif
+allow_experimental_apis = true
 deps += ['hash', 'vhost']
 cflags += '-Wno-deprecated-declarations'
 sources = files(
Signature · 1 line
2.20.1
#9 lore raw

PATCH 7/7 examples/vhost: block attempts to use owned ports

In reply to Stephen Hemminger

If a ethdev port is in use for a sub device, then it should not be allowed in the portmask of application.

Fixes 5b7ba31148a8 ("ethdev: add port ownership")
---
 examples/vhost/main.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
examples/vhost/main.c +1 -1
index ab649bf147e1..d759581f1904 100644
@@ -669,7 +669,7 @@ static unsigned check_ports_num(unsigned nb_ports)
 	}
 
 	for (portid = 0; portid < num_ports; portid ++) {
-		if (!rte_eth_dev_is_valid_port(ports[portid])) {
+		if (!rte_eth_dev_is_owned_by(portid, RTE_ETH_DEV_NO_OWNER)) {
 			RTE_LOG(INFO, VHOST_PORT,
 				"\nSpecified port ID(%u) is not valid\n",
 				ports[portid]);
Signature · 1 line
2.20.1
#10 lore raw

v2 cover check for owned ports in portmask

In reply to Stephen Hemminger

The failsafe, bonding, and netvsc PMD use eth dev port ownership to control sub devices. These sub devices are hidden in the normal iteration over ports but some applications use a direct port mask. In these cases, user may still (incorrectly) try to directly use the sub devices.

This patch set introduces a helper function to check if a port is under ownership. And the follow on patches use that.

An alternative proposed in earlier discussion is to have each application try and own the port. But this solution would be more invasive, and does not handle the case if secondary process exits prematurely without releasing ownership. It makes more sense to keep the concept of ownership as strictly part of the device model and not part of the application layer.

v2 - rename the helper function and use rte_eth_dev_owner_get

Stephen Hemminger (6):
rte_ethdev: add function to check if device is owned examples/l2fwd-cat: block attempts to use owned ports examples/l3fwd: block attempts to use owned ports examples/l3fwd-acl: block attempts to use owned ports examples/l3fwd-power: block attempts to use owned ports examples/tep_termination: block attempts to use owned ports

 examples/l2fwd-cat/Makefile              |  2 ++
 examples/l2fwd-cat/l2fwd-cat.c           |  3 +++
 examples/l2fwd-cat/meson.build           |  1 +
 examples/l3fwd-acl/Makefile              |  3 +++
 examples/l3fwd-acl/main.c                |  4 ++++
 examples/l3fwd-acl/meson.build           |  1 +
 examples/l3fwd-power/main.c              |  4 ++++
 examples/l3fwd/Makefile                  |  3 +++
 examples/l3fwd/main.c                    |  5 +++++
 examples/l3fwd/meson.build               |  1 +
 examples/tep_termination/Makefile        |  2 ++
 examples/tep_termination/main.c          |  2 +-
 examples/tep_termination/meson.build     |  1 +
 lib/librte_ethdev/rte_ethdev.c           | 12 ++++++++++++
 lib/librte_ethdev/rte_ethdev.h           | 13 +++++++++++++
 lib/librte_ethdev/rte_ethdev_version.map |  3 +++
 16 files changed, 59 insertions(+), 1 deletion(-)
Signature · 1 line
2.20.1
#11 lore raw

PATCH v2 1/6 rte_ethdev: add function to check if device is owned

This is a simple helper function to check if device is owned (being used as a sub-device). It is more convienent than having applications call rte_eth_dev_owner_get and check the result.

---
v2 - rename the helper function and use rte_eth_dev_owner_get

 lib/librte_ethdev/rte_ethdev.c           | 12 ++++++++++++
 lib/librte_ethdev/rte_ethdev.h           | 13 +++++++++++++
 lib/librte_ethdev/rte_ethdev_version.map |  3 +++
 3 files changed, 28 insertions(+)
lib/librte_ethdev/rte_ethdev.c +12 -0
index 774c721b3484..6e2d0110f65e 100644
@@ -737,6 +737,18 @@ rte_eth_dev_owner_get(const uint16_t port_id, struct rte_eth_dev_owner *owner)
 	return ret;
 }
 
+int
+rte_eth_dev_is_owned(uint16_t port_id)
+{
+	struct rte_eth_dev_owner owner;
+	int ret;
+
+	ret = rte_eth_dev_owner_get(port_id, &owner);
+	if (ret == 0)
+		ret = (owner.id != RTE_ETH_DEV_NO_OWNER);
+	return ret;
+}
+
 int
 rte_eth_dev_socket_id(uint16_t port_id)
 {
lib/librte_ethdev/rte_ethdev.h +13 -0
index d1a593ad112a..3a8a4bcde4b7 100644
@@ -1693,6 +1693,19 @@ __rte_experimental
 int rte_eth_dev_owner_get(const uint16_t port_id,
 		struct rte_eth_dev_owner *owner);
 
+/**
+ * Check if port_id is part of a sub-device
+ *
+ * @param port_id
+ *   The port identifier of the Ethernet device
+ * @return
+ *   - 1 if device is associated with an owner
+ *   - 0 if port is not owned
+ *   - negative errno value on error.
+ */
+__rte_experimental
+int rte_eth_dev_is_owned(uint16_t port_id);
+
 /**
  * Get the number of ports which are usable for the application.
  *
lib/librte_ethdev/rte_ethdev_version.map +3 -0
index 3f32fdecf722..435df2dba149 100644
@@ -230,4 +230,7 @@ EXPERIMENTAL {
 
 	# added in 20.02
 	rte_flow_dev_dump;
+
+	# added in 20.05
+	rte_eth_dev_is_owned;
 };
Signature · 1 line
2.20.1
#12 lore raw

Re: [dpdk-dev] [PATCH v2 1/6] rte_ethdev: add function to check if device is owned

16/03/2020 17:09, Stephen Hemminger:

Quoted 2 lines

This is a simple helper function to check if device is owned (being used as a sub-device).

I would suggest not restricting ownership to sub-device case.

Quoted 2 lines

It is more convienent than having
applications call rte_eth_dev_owner_get and check the result.

Yes it is more convenient but I don't like adding such simple wrapper.

I propose to extend rte_eth_dev_owner_get() behaviour:
if the owner pointer is NULL, the function returns 0 only if an owner (not RTE_ETH_DEV_NO_OWNER) is found.

So instead of using your wrapper:

	if (rte_eth_dev_is_owned(port_id))
you can use:
	if (rte_eth_dev_owner_get(port_id, NULL) == 0)

[...]

Quoted 11 lines
lib/librte_ethdev/rte_ethdev.c +11 -0
+int
+rte_eth_dev_is_owned(uint16_t port_id)
+{
+	struct rte_eth_dev_owner owner;
+	int ret;
+
+	ret = rte_eth_dev_owner_get(port_id, &owner);
+	if (ret == 0)
+		ret = (owner.id != RTE_ETH_DEV_NO_OWNER);
+	return ret;
+}
#13 lore raw

Re: [dpdk-dev] [PATCH v2 1/6] rte_ethdev: add function to check if device is owned

Quoted 11 lines · On Wed, 01 Apr 2020 23:42:44 +0200 Thomas Monjalon <[email protected]> wrote:

16/03/2020 17:09, Stephen Hemminger:

Quoted 2 lines

This is a simple helper function to check if device is owned (being used as a sub-device).

I would suggest not restricting ownership to sub-device case.

Quoted 2 lines

It is more convienent than having
applications call rte_eth_dev_owner_get and check the result.

Yes it is more convenient but I don't like adding such simple wrapper.

Quoted 8 lines

I propose to extend rte_eth_dev_owner_get() behaviour:
if the owner pointer is NULL, the function returns 0 only if an owner (not RTE_ETH_DEV_NO_OWNER) is found.

So instead of using your wrapper:

	if (rte_eth_dev_is_owned(port_id))
you can use:
	if (rte_eth_dev_owner_get(port_id, NULL) == 0)

That is not how rte_eth_dev_owner_get works now.
Passing NULL to it would crash.

And if devices is not owned rte_eth_dev_owner_get() returns 0 and owner is set to a magic value. We could change the ABI for this since it is marked experimental. But that seems more risky.

#14 lore raw

Re: [dpdk-dev] [PATCH v2 1/6] rte_ethdev: add function to check if device is owned

02/04/2020 00:24, Stephen Hemminger:

Quoted 18 lines
Quoted 15 lines · On Wed, 01 Apr 2020 23:42:44 +0200 Thomas Monjalon <[email protected]> wrote:

16/03/2020 17:09, Stephen Hemminger:

Quoted 2 lines

This is a simple helper function to check if device is owned (being used as a sub-device).

I would suggest not restricting ownership to sub-device case.

Quoted 2 lines

It is more convienent than having
applications call rte_eth_dev_owner_get and check the result.

Yes it is more convenient but I don't like adding such simple wrapper.

I propose to extend rte_eth_dev_owner_get() behaviour:
if the owner pointer is NULL, the function returns 0 only if an owner (not RTE_ETH_DEV_NO_OWNER) is found.

Quoted 7 lines
Quoted 4 lines

So instead of using your wrapper:

	if (rte_eth_dev_is_owned(port_id))
you can use:
	if (rte_eth_dev_owner_get(port_id, NULL) == 0)

That is not how rte_eth_dev_owner_get works now.
Passing NULL to it would crash.

Indeed this is why I am proposing to extend it
and fill this case.

Quoted 3 lines

And if devices is not owned rte_eth_dev_owner_get() returns 0 and owner is set to a magic value. We could change the ABI for this since it is marked experimental. But that seems more risky.

This is not an ABI change.
Just manage a case which was crashing previously.

#15 lore raw

PATCH v2 2/6 examples/l2fwd-cat: block attempts to use owned ports

In reply to Stephen Hemminger

If a ethdev port is in use for a sub device, then it should not be allowed in the portmask of application.

Fixes 5b7ba31148a8 ("ethdev: add port ownership")
---
 examples/l2fwd-cat/Makefile    | 2 ++
 examples/l2fwd-cat/l2fwd-cat.c | 3 +++
 examples/l2fwd-cat/meson.build | 1 +
 3 files changed, 6 insertions(+)
examples/l2fwd-cat/Makefile +2 -0
index b0e53c37e8d8..aa19347c1545 100644
@@ -20,6 +20,7 @@ static: build/$(APP)-static
 PKGCONF ?= pkg-config
 
 PC_FILE := $(shell $(PKGCONF) --path libdpdk 2>/dev/null)
+CFLAGS += -DALLOW_EXPERIMENTAL_API
 CFLAGS += -O3 $(shell $(PKGCONF) --cflags libdpdk)
 LDFLAGS_SHARED = $(shell $(PKGCONF) --libs libdpdk)
 LDFLAGS_STATIC = -Wl,-Bstatic $(shell $(PKGCONF) --static --libs libdpdk)
@@ -65,6 +66,7 @@ endif
 
 EXTRA_CFLAGS += -O3 -g -Wfatal-errors
 
+CFLAGS += -DALLOW_EXPERIMENTAL_API
 CFLAGS += -I$(PQOS_INSTALL_PATH)/../include
 
 LDLIBS += -L$(PQOS_INSTALL_PATH)
examples/l2fwd-cat/l2fwd-cat.c +3 -0
index 6838f288c621..e89b3568bc86 100644
@@ -42,6 +42,9 @@ port_init(uint16_t port, struct rte_mempool *mbuf_pool)
 	if (!rte_eth_dev_is_valid_port(port))
 		return -1;
 
+	if (rte_eth_dev_is_owned(portid))
+		return -1;
+
 	/* Configure the Ethernet device. */
 	retval = rte_eth_dev_configure(port, rx_rings, tx_rings, &port_conf);
 	if (retval != 0)
examples/l2fwd-cat/meson.build +1 -0
index 4e2777a03358..5c062c1354e3 100644
@@ -6,6 +6,7 @@
 # To build this example as a standalone application with an already-installed
 # DPDK instance, use 'make'
 
+allow_experimental_apis = true
 pqos = cc.find_library('pqos', required: false)
 build = pqos.found()
 ext_deps += pqos
Signature · 1 line
2.20.1
#16 lore raw

PATCH v2 3/6 examples/l3fwd: block attempts to use owned ports

In reply to Stephen Hemminger

If a ethdev port is in use for a sub device, then it should not be allowed in the portmask of l3fwd.

Fixes 5b7ba31148a8 ("ethdev: add port ownership")
---
 examples/l3fwd/Makefile    | 3 +++
 examples/l3fwd/main.c      | 5 +++++
 examples/l3fwd/meson.build | 1 +
 3 files changed, 9 insertions(+)
examples/l3fwd/Makefile +3 -0
index 59a110d12a49..59cd00abe6e9 100644
@@ -25,6 +25,8 @@ CFLAGS += -O3 $(shell $(PKGCONF) --cflags libdpdk)
 LDFLAGS_SHARED = $(shell $(PKGCONF) --libs libdpdk)
 LDFLAGS_STATIC = -Wl,-Bstatic $(shell $(PKGCONF) --static --libs libdpdk)
 
+CFLAGS += -DALLOW_EXPERIMENTAL_API
+
 build/$(APP)-shared: $(SRCS-y) Makefile $(PC_FILE) | build
 	$(CC) $(CFLAGS) $(SRCS-y) -o $@ $(LDFLAGS) $(LDFLAGS_SHARED)
 
@@ -50,6 +52,7 @@ RTE_TARGET ?= $(notdir $(abspath $(dir $(firstword $(wildcard $(RTE_SDK)/*/.conf
 
 include $(RTE_SDK)/mk/rte.vars.mk
 
+CFLAGS += -DALLOW_EXPERIMENTAL_API
 CFLAGS += -I$(SRCDIR)
 CFLAGS += -O3 $(USER_FLAGS)
 CFLAGS += $(WERROR_FLAGS)
examples/l3fwd/main.c +5 -0
index dda430d68a20..ab717e3bfd2e 100644
@@ -221,6 +221,11 @@ check_port_config(void)
 			printf("port %u is not present on the board\n", portid);
 			return -1;
 		}
+
+		if (rte_eth_dev_is_owned(portid)) {
+			printf("port %u is already in use\n", portid);
+			return -1;
+		}
 	}
 	return 0;
 }
examples/l3fwd/meson.build +1 -0
index ebed3b518c4b..7d72b1b365ae 100644
@@ -6,6 +6,7 @@
 # To build this example as a standalone application with an already-installed
 # DPDK instance, use 'make'
 
+allow_experimental_apis = true
 deps += ['hash', 'lpm', 'eventdev']
 sources = files(
 	'l3fwd_em.c', 'l3fwd_lpm.c', 'l3fwd_event.c',
Signature · 1 line
2.20.1
#17 lore raw

PATCH v2 4/6 examples/l3fwd-acl: block attempts to use owned ports

In reply to Stephen Hemminger

If a ethdev port is in use for a sub device, then it should not be allowed in the portmask of application.

Fixes 5b7ba31148a8 ("ethdev: add port ownership")
---
 examples/l3fwd-acl/Makefile    | 3 +++
 examples/l3fwd-acl/main.c      | 4 ++++
 examples/l3fwd-acl/meson.build | 1 +
 3 files changed, 8 insertions(+)
examples/l3fwd-acl/Makefile +3 -0
index d9909584b1c6..ff14b4fdd593 100644
@@ -20,7 +20,9 @@ static: build/$(APP)-static
 PKGCONF ?= pkg-config
 
 PC_FILE := $(shell $(PKGCONF) --path libdpdk 2>/dev/null)
+CFLAGS += -DALLOW_EXPERIMENTAL_API
 CFLAGS += -O3 $(shell $(PKGCONF) --cflags libdpdk)
+
 LDFLAGS_SHARED = $(shell $(PKGCONF) --libs libdpdk)
 LDFLAGS_STATIC = -Wl,-Bstatic $(shell $(PKGCONF) --static --libs libdpdk)
 
@@ -51,6 +53,7 @@ include $(RTE_SDK)/mk/rte.vars.mk
 
 CFLAGS += -O3
 CFLAGS += $(WERROR_FLAGS)
+CFLAGS += -DALLOW_EXPERIMENTAL_API
 
 # workaround for a gcc bug with noreturn attribute
 # http://gcc.gnu.org/bugzilla/show_bug.cgi?id=12603
examples/l3fwd-acl/main.c +4 -0
index fa92a2829740..d9bee9ade323 100644
@@ -1470,6 +1470,10 @@ check_port_config(void)
 			printf("port %u is not present on the board\n", portid);
 			return -1;
 		}
+		if (rte_eth_dev_is_owned(portid)) {
+			printf("port %u is already in use\n", portid);
+			return -1;
+		}
 	}
 	return 0;
 }
examples/l3fwd-acl/meson.build +1 -0
index 7096e00c1073..6fa468b3aa9c 100644
@@ -6,6 +6,7 @@
 # To build this example as a standalone application with an already-installed
 # DPDK instance, use 'make'
 
+allow_experimental_apis = true
 deps += ['acl', 'lpm', 'hash']
 sources = files(
 	'main.c'
Signature · 1 line
2.20.1
#18 lore raw

PATCH v2 5/6 examples/l3fwd-power: block attempts to use owned ports

In reply to Stephen Hemminger

If a ethdev port is in use for a sub device, then it should not be allowed in the portmask of l3fwd.

Fixes 5b7ba31148a8 ("ethdev: add port ownership")
---
 examples/l3fwd-power/main.c | 4 ++++
 1 file changed, 4 insertions(+)
examples/l3fwd-power/main.c +4 -0
index c7fe0ec03495..60f5ddd8fb9e 100644
@@ -1396,6 +1396,10 @@ check_port_config(void)
 								portid);
 			return -1;
 		}
+		if (rte_eth_dev_is_owned(portid)) {
+			printf("port %u is already in use\n", portid);
+			return -1;
+		}
 	}
 	return 0;
 }
Signature · 1 line
2.20.1
#19 lore raw

PATCH v2 6/6 examples/tep_termination: block attempts to use owned ports

In reply to Stephen Hemminger

If a ethdev port is in use for a sub device, then it should not be allowed in the portmask of application.

Fixes 5b7ba31148a8 ("ethdev: add port ownership")
---
 examples/tep_termination/Makefile    | 2 ++
 examples/tep_termination/main.c      | 2 +-
 examples/tep_termination/meson.build | 1 +
 3 files changed, 4 insertions(+), 1 deletion(-)
examples/tep_termination/Makefile +2 -0
index 645112498d33..f55331fb24c2 100644
@@ -26,6 +26,7 @@ CFLAGS += -O3 $(shell $(PKGCONF) --cflags libdpdk)
 LDFLAGS_SHARED = $(shell $(PKGCONF) --libs libdpdk)
 LDFLAGS_STATIC = -Wl,-Bstatic $(shell $(PKGCONF) --static --libs libdpdk)
 
+CFLAGS += -DALLOW_EXPERIMENTAL_API
 CFLAGS += -Wno-deprecated-declarations
 
 build/$(APP)-shared: $(SRCS-y) Makefile $(PC_FILE) | build
@@ -61,6 +62,7 @@ endif
 CFLAGS += -O3
 CFLAGS += $(WERROR_FLAGS)
 CFLAGS += -Wno-deprecated-declarations
+CFLAGS += -DALLOW_EXPERIMENTAL_API
 
 include $(RTE_SDK)/mk/rte.extapp.mk
 endif
examples/tep_termination/main.c +1 -1
index ab956ad7ce96..96c2d934cd09 100644
@@ -507,7 +507,7 @@ check_ports_num(unsigned max_nb_ports)
 	}
 
 	for (portid = 0; portid < nb_ports; portid++) {
-		if (!rte_eth_dev_is_valid_port(ports[portid])) {
+		if (rte_eth_dev_is_owned(portid) != 0) {
 			RTE_LOG(INFO, VHOST_PORT,
 				"\nSpecified port ID(%u) is not valid\n",
 				ports[portid]);
examples/tep_termination/meson.build +1 -0
index f65d689802a3..ddadf1c2aede 100644
@@ -9,6 +9,7 @@
 if not is_linux
 	build = false
 endif
+allow_experimental_apis = true
 deps += ['hash', 'vhost']
 cflags += '-Wno-deprecated-declarations'
 sources = files(
Signature · 1 line
2.20.1
#20 lore raw

Re: [dpdk-dev] [PATCH v2 0/6] check for owned ports in portmask

In reply to Stephen Hemminger

From: Stephen Hemminger

Quoted 8 lines

The failsafe, bonding, and netvsc PMD use eth dev port ownership to control sub devices. These sub devices are hidden in the normal iteration over ports but some applications use a direct port mask. In these cases, user may still (incorrectly) try to directly use the sub devices.

This patch set introduces a helper function to check if a port is under ownership. And the follow on patches use that.

Quoted 7 lines

An alternative proposed in earlier discussion is to have each application try and own the port. But this solution would be more invasive, and does not handle the case if secondary process exits prematurely without releasing ownership. It makes more sense to keep the concept of ownership as strictly part of the device model and not part of the application layer.

v2 - rename the helper function and use rte_eth_dev_owner_get

Quoted 28 lines

Stephen Hemminger (6):
rte_ethdev: add function to check if device is owned examples/l2fwd-cat: block attempts to use owned ports examples/l3fwd: block attempts to use owned ports examples/l3fwd-acl: block attempts to use owned ports examples/l3fwd-power: block attempts to use owned ports examples/tep_termination: block attempts to use owned ports

 examples/l2fwd-cat/Makefile              |  2 ++
 examples/l2fwd-cat/l2fwd-cat.c           |  3 +++
 examples/l2fwd-cat/meson.build           |  1 +
 examples/l3fwd-acl/Makefile              |  3 +++
 examples/l3fwd-acl/main.c                |  4 ++++
 examples/l3fwd-acl/meson.build           |  1 +
 examples/l3fwd-power/main.c              |  4 ++++
 examples/l3fwd/Makefile                  |  3 +++
 examples/l3fwd/main.c                    |  5 +++++
 examples/l3fwd/meson.build               |  1 +
 examples/tep_termination/Makefile        |  2 ++
 examples/tep_termination/main.c          |  2 +-
 examples/tep_termination/meson.build     |  1 +
 lib/librte_ethdev/rte_ethdev.c           | 12 ++++++++++++
 lib/librte_ethdev/rte_ethdev.h           | 13 +++++++++++++
 lib/librte_ethdev/rte_ethdev_version.map |  3 +++
 16 files changed, 59 insertions(+), 1 deletion(-)

--
2.20.1

#21 lore raw

v3 cover checking for owned ports in portmask

In reply to Stephen Hemminger

The failsafe, bonding, and netvsc PMD use eth dev port ownership to control sub devices. These sub devices are hidden in the normal iteration over ports but some applications use a direct port mask. In these cases, user may still (incorrectly) try to directly use the sub devices.

This patch set changes the rte_eth_dev_owner_get function so that it can be used to test for any ownership. And the follow on patches use that.

An alternative proposed in earlier discussion is to have each application try and own the port. But this solution would be more invasive, and does not handle the case if secondary process exits prematurely without releasing ownership. It makes more sense to keep the concept of ownership as strictly part of the device model and not part of the application layer.

v3 - use rte_eth_dev_owner_get directly, and change the function return
rework port setup in tep_termination example
v2 - rename the helper function and use rte_eth_dev_owner_get

Stephen Hemminger (6):
rte_ethdev: change rte_eth_dev_owner_get to return error if unowned examples/l2fwd-cat: make applicaton aware of port ownership examples/l3fwd: make applicaton aware of port ownership examples/l3fwd-acl: make applicaton aware of port ownership examples/l3fwd-power: make applicaton aware of port ownership examples/tep_termination: rework the port setup logic

 drivers/net/failsafe/failsafe_eal.c  |  4 +-
 drivers/net/netvsc/hn_vf.c           | 11 ++---
 examples/l2fwd-cat/Makefile          |  2 +
 examples/l2fwd-cat/l2fwd-cat.c       |  3 ++
 examples/l2fwd-cat/meson.build       |  1 +
 examples/l3fwd-acl/Makefile          |  3 ++
 examples/l3fwd-acl/main.c            |  4 ++
 examples/l3fwd-acl/meson.build       |  1 +
 examples/l3fwd-power/main.c          |  4 ++
 examples/l3fwd/Makefile              |  3 ++
 examples/l3fwd/main.c                |  5 ++
 examples/l3fwd/meson.build           |  1 +
 examples/tep_termination/Makefile    |  2 +
 examples/tep_termination/main.c      | 70 ++++++++++------------------
 examples/tep_termination/meson.build |  1 +
 lib/librte_ethdev/rte_ethdev.c       |  4 +-
 lib/librte_ethdev/rte_ethdev.h       |  6 ++-
 17 files changed, 69 insertions(+), 56 deletions(-)
Signature · 1 line
2.20.1
#22 lore raw

PATCH v3 1/6 rte_ethdev: change rte_eth_dev_owner_get to return error if unowned

For applications that want to check if device is owned, change the return value of rte_eth_dev_owner_get to return -ENOENT if there is no owner for the device.

Change the two drivers (failsafe and netvsc) that are using this experimental API to handle the new return value.

---
 drivers/net/failsafe/failsafe_eal.c |  4 ++--
 drivers/net/netvsc/hn_vf.c          | 11 +++++------
 lib/librte_ethdev/rte_ethdev.c      |  4 +++-
 lib/librte_ethdev/rte_ethdev.h      |  6 ++++--
 4 files changed, 14 insertions(+), 11 deletions(-)
drivers/net/failsafe/failsafe_eal.c +2 -2
index b9fc508673f2..8a1b5e068840 100644
@@ -63,8 +63,8 @@ fs_bus_init(struct rte_eth_dev *dev)
 			 * The NEW callback tried to take ownership, check
 			 * whether it succeed or didn't.
 			 */
-			rte_eth_dev_owner_get(pid, &pid_owner);
-			if (pid_owner.id != PRIV(dev)->my_owner.id) {
+			if (rte_eth_dev_owner_get(pid, &pid_owner) == 0 &&
+			    pid_owner.id != PRIV(dev)->my_owner.id) {
 				INFO("sub_device %d owner(%s_%016"PRIX64") is not my,"
 				     " owner(%s_%016"PRIX64"), will try again later",
 				     i, pid_owner.name, pid_owner.id,
drivers/net/netvsc/hn_vf.c +5 -6
index 7a3734cadfa4..b56f60b28a62 100644
@@ -63,16 +63,15 @@ static int hn_vf_attach(struct hn_data *hv, uint16_t port_id)
 	}
 
 	ret = rte_eth_dev_owner_get(port_id, &owner);
-	if (ret < 0) {
-		PMD_DRV_LOG(ERR, "Can not find owner for port %d", port_id);
-		return ret;
-	}
-
-	if (owner.id != RTE_ETH_DEV_NO_OWNER) {
+	if (ret == 0) {
 		PMD_DRV_LOG(ERR, "Port %u already owned by other device %s",
 			    port_id, owner.name);
 		return -EBUSY;
 	}
+	if (ret != -ENOENT) {
+		PMD_DRV_LOG(ERR, "Can not find owner for port %d", port_id);
+		return ret;
+	}
 
 	ret = rte_eth_dev_owner_set(port_id, &hv->owner);
 	if (ret < 0) {
lib/librte_ethdev/rte_ethdev.c +3 -1
index 0854ef883270..0c4645cf9832 100644
@@ -729,7 +729,9 @@ rte_eth_dev_owner_get(const uint16_t port_id, struct rte_eth_dev_owner *owner)
 		RTE_ETHDEV_LOG(ERR, "Port id %"PRIu16" is not allocated\n",
 			port_id);
 		ret = -ENODEV;
-	} else {
+	} else if (ethdev->data->owner.id == RTE_ETH_DEV_NO_OWNER) {
+		ret = -ENOENT;
+	} else if (owner != NULL) {
 		rte_memcpy(owner, &ethdev->data->owner, sizeof(*owner));
 	}
 
lib/librte_ethdev/rte_ethdev.h +4 -2
index d1a593ad112a..0e59d39de54c 100644
@@ -1685,9 +1685,11 @@ int rte_eth_dev_owner_delete(const uint64_t owner_id);
  * @param	port_id
  *  The port identifier.
  * @param	owner
- *  The owner structure pointer to fill.
+ *  The owner structure pointer to fill (optional can be NULL)
  * @return
- *  0 on success, negative errno value on error..
+ *  0 on success,
+ *  -ENODEV if port_id is not valid
+ *  -ENOENT if the device is ownerless.
  */
 __rte_experimental
 int rte_eth_dev_owner_get(const uint16_t port_id,
Signature · 1 line
2.20.1
#23 lore raw

Re: [dpdk-dev] [PATCH v3 1/6] rte_ethdev: change rte_eth_dev_owner_get to return error if unowned

Quoted 8 lines · Stephen Hemminger

For applications that want to check if device is owned, change the return value of rte_eth_dev_owner_get to return -ENOENT if there is no owner for the device.

Change the two drivers (failsafe and netvsc) that are using this experimental API to handle the new return value.

<...>

lib/librte_ethdev/rte_ethdev.h int rte_eth_dev_owner_delete(const uint64_t owner_id); · line 1692

Quoted 11 lines
lib/librte_ethdev/rte_ethdev.h +4 -2
@@ -1685,9 +1685,11 @@ int rte_eth_dev_owner_delete(const uint64_t owner_id);
  * @param	port_id
  *  The port identifier.
  * @param	owner
- *  The owner structure pointer to fill.
+ *  The owner structure pointer to fill (optional can be NULL)
  * @return
- *  0 on success, negative errno value on error..
+ *  0 on success,
+ *  -ENODEV if port_id is not valid
+ *  -ENOENT if the device is ownerless.

Not sure if port not having an owner is error, I understand the motivation but what about send a special value, like 1, instead of error:
0: get owner success, port owned
1: get owner success, port not owned
-ENODEV: get owner failed, port_id is not valid

#24 lore raw

PATCH v3 2/6 examples/l2fwd-cat: make applicaton aware of port ownership

In reply to Stephen Hemminger

If a ethdev port is in use for a sub device, then it should not be allowed in the portmask of application.

Fixes 5b7ba31148a8 ("ethdev: add port ownership")
---
 examples/l2fwd-cat/Makefile    | 2 ++
 examples/l2fwd-cat/l2fwd-cat.c | 3 +++
 examples/l2fwd-cat/meson.build | 1 +
 3 files changed, 6 insertions(+)
examples/l2fwd-cat/Makefile +2 -0
index b0e53c37e8d8..aa19347c1545 100644
@@ -20,6 +20,7 @@ static: build/$(APP)-static
 PKGCONF ?= pkg-config
 
 PC_FILE := $(shell $(PKGCONF) --path libdpdk 2>/dev/null)
+CFLAGS += -DALLOW_EXPERIMENTAL_API
 CFLAGS += -O3 $(shell $(PKGCONF) --cflags libdpdk)
 LDFLAGS_SHARED = $(shell $(PKGCONF) --libs libdpdk)
 LDFLAGS_STATIC = -Wl,-Bstatic $(shell $(PKGCONF) --static --libs libdpdk)
@@ -65,6 +66,7 @@ endif
 
 EXTRA_CFLAGS += -O3 -g -Wfatal-errors
 
+CFLAGS += -DALLOW_EXPERIMENTAL_API
 CFLAGS += -I$(PQOS_INSTALL_PATH)/../include
 
 LDLIBS += -L$(PQOS_INSTALL_PATH)
examples/l2fwd-cat/l2fwd-cat.c +3 -0
index 6838f288c621..e370efeaf1f8 100644
@@ -42,6 +42,9 @@ port_init(uint16_t port, struct rte_mempool *mbuf_pool)
 	if (!rte_eth_dev_is_valid_port(port))
 		return -1;
 
+	if (rte_eth_dev_owner_get(port, NULL) == 0)
+		return -1;
+
 	/* Configure the Ethernet device. */
 	retval = rte_eth_dev_configure(port, rx_rings, tx_rings, &port_conf);
 	if (retval != 0)
examples/l2fwd-cat/meson.build +1 -0
index 4e2777a03358..5c062c1354e3 100644
@@ -6,6 +6,7 @@
 # To build this example as a standalone application with an already-installed
 # DPDK instance, use 'make'
 
+allow_experimental_apis = true
 pqos = cc.find_library('pqos', required: false)
 build = pqos.found()
 ext_deps += pqos
Signature · 1 line
2.20.1
#25 lore raw

Re: [dpdk-dev] [dpdk-stable] [PATCH v3 2/6] examples/l2fwd-cat: make applicaton aware of port ownership

Quoted 7 lines · Stephen Hemminger

If a ethdev port is in use for a sub device, then it should not be allowed in the portmask of application.

Fixes 5b7ba31148a8 ("ethdev: add port ownership")

<...>

examples/l2fwd-cat/l2fwd-cat.c port_init(uint16_t port, struct rte_mempool *mbuf_pool) · line 46

Quoted 6 lines
examples/l2fwd-cat/l2fwd-cat.c +2 -0
@@ -42,6 +42,9 @@ port_init(uint16_t port, struct rte_mempool *mbuf_pool)
 	if (!rte_eth_dev_is_valid_port(port))
 		return -1;
 
+	if (rte_eth_dev_owner_get(port, NULL) == 0)
+		return -1;

'port_init' is called by the loop:

RTE_ETH_FOREACH_DEV(portid)

    if (port_init(portid, mbuf_pool) != 0)

And 'RTE_ETH_FOREACH_DEV' already iterates only on ports not owned, so this check is redundant.

#26 lore raw

PATCH v3 3/6 examples/l3fwd: make applicaton aware of port ownership

In reply to Stephen Hemminger

If a ethdev port is in use for a sub device, then it should not be allowed in the portmask of l3fwd.

Fixes 5b7ba31148a8 ("ethdev: add port ownership")
---
 examples/l3fwd/Makefile    | 3 +++
 examples/l3fwd/main.c      | 5 +++++
 examples/l3fwd/meson.build | 1 +
 3 files changed, 9 insertions(+)
examples/l3fwd/Makefile +3 -0
index 59a110d12a49..59cd00abe6e9 100644
@@ -25,6 +25,8 @@ CFLAGS += -O3 $(shell $(PKGCONF) --cflags libdpdk)
 LDFLAGS_SHARED = $(shell $(PKGCONF) --libs libdpdk)
 LDFLAGS_STATIC = -Wl,-Bstatic $(shell $(PKGCONF) --static --libs libdpdk)
 
+CFLAGS += -DALLOW_EXPERIMENTAL_API
+
 build/$(APP)-shared: $(SRCS-y) Makefile $(PC_FILE) | build
 	$(CC) $(CFLAGS) $(SRCS-y) -o $@ $(LDFLAGS) $(LDFLAGS_SHARED)
 
@@ -50,6 +52,7 @@ RTE_TARGET ?= $(notdir $(abspath $(dir $(firstword $(wildcard $(RTE_SDK)/*/.conf
 
 include $(RTE_SDK)/mk/rte.vars.mk
 
+CFLAGS += -DALLOW_EXPERIMENTAL_API
 CFLAGS += -I$(SRCDIR)
 CFLAGS += -O3 $(USER_FLAGS)
 CFLAGS += $(WERROR_FLAGS)
examples/l3fwd/main.c +5 -0
index dda430d68a20..110d4cc109d1 100644
@@ -221,6 +221,11 @@ check_port_config(void)
 			printf("port %u is not present on the board\n", portid);
 			return -1;
 		}
+
+		if (rte_eth_dev_owner_get(portid, NULL) == 0) {
+			printf("port %u is already in use\n", portid);
+			return -1;
+		}
 	}
 	return 0;
 }
examples/l3fwd/meson.build +1 -0
index ebed3b518c4b..7d72b1b365ae 100644
@@ -6,6 +6,7 @@
 # To build this example as a standalone application with an already-installed
 # DPDK instance, use 'make'
 
+allow_experimental_apis = true
 deps += ['hash', 'lpm', 'eventdev']
 sources = files(
 	'l3fwd_em.c', 'l3fwd_lpm.c', 'l3fwd_event.c',
Signature · 1 line
2.20.1
#27 lore raw

PATCH v3 4/6 examples/l3fwd-acl: make applicaton aware of port ownership

In reply to Stephen Hemminger

If a ethdev port is in use for a sub device, then it should not be allowed in the portmask of application.

Fixes 5b7ba31148a8 ("ethdev: add port ownership")
---
 examples/l3fwd-acl/Makefile    | 3 +++
 examples/l3fwd-acl/main.c      | 4 ++++
 examples/l3fwd-acl/meson.build | 1 +
 3 files changed, 8 insertions(+)
examples/l3fwd-acl/Makefile +3 -0
index d9909584b1c6..ff14b4fdd593 100644
@@ -20,7 +20,9 @@ static: build/$(APP)-static
 PKGCONF ?= pkg-config
 
 PC_FILE := $(shell $(PKGCONF) --path libdpdk 2>/dev/null)
+CFLAGS += -DALLOW_EXPERIMENTAL_API
 CFLAGS += -O3 $(shell $(PKGCONF) --cflags libdpdk)
+
 LDFLAGS_SHARED = $(shell $(PKGCONF) --libs libdpdk)
 LDFLAGS_STATIC = -Wl,-Bstatic $(shell $(PKGCONF) --static --libs libdpdk)
 
@@ -51,6 +53,7 @@ include $(RTE_SDK)/mk/rte.vars.mk
 
 CFLAGS += -O3
 CFLAGS += $(WERROR_FLAGS)
+CFLAGS += -DALLOW_EXPERIMENTAL_API
 
 # workaround for a gcc bug with noreturn attribute
 # http://gcc.gnu.org/bugzilla/show_bug.cgi?id=12603
examples/l3fwd-acl/main.c +4 -0
index fa92a2829740..dbdea8db241a 100644
@@ -1470,6 +1470,10 @@ check_port_config(void)
 			printf("port %u is not present on the board\n", portid);
 			return -1;
 		}
+		if (rte_eth_dev_owner_get(portid, NULL) == 0) {
+			printf("port %u is already in use\n", portid);
+			return -1;
+		}
 	}
 	return 0;
 }
examples/l3fwd-acl/meson.build +1 -0
index 7096e00c1073..6fa468b3aa9c 100644
@@ -6,6 +6,7 @@
 # To build this example as a standalone application with an already-installed
 # DPDK instance, use 'make'
 
+allow_experimental_apis = true
 deps += ['acl', 'lpm', 'hash']
 sources = files(
 	'main.c'
Signature · 1 line
2.20.1
#28 lore raw

PATCH v3 5/6 examples/l3fwd-power: make applicaton aware of port ownership

In reply to Stephen Hemminger

If a ethdev port is in use for a sub device, then it should not be allowed in the portmask of l3fwd.

Fixes 5b7ba31148a8 ("ethdev: add port ownership")
---
 examples/l3fwd-power/main.c | 4 ++++
 1 file changed, 4 insertions(+)
examples/l3fwd-power/main.c +4 -0
index c7fe0ec03495..186d6ac8e271 100644
@@ -1396,6 +1396,10 @@ check_port_config(void)
 								portid);
 			return -1;
 		}
+		if (rte_eth_dev_owner_get(portid, NULL) == 0) {
+			printf("port %u is already in use\n", portid);
+			return -1;
+		}
 	}
 	return 0;
 }
Signature · 1 line
2.20.1
#29 lore raw

PATCH v3 6/6 examples/tep_termination: rework the port setup logic

In reply to Stephen Hemminger

The handling of ports in this application had many problems. It was checking for things that can never happen with current DPDK library (like rte_ethdev_avail_count() >= RTE_MAX_ETHPORTS) and it was not checking if the port was owned and should not be used directly. Also the variable nb_ports was used as both local and global variable.

Fix by rewriting the initialization logic to iterate over valid ports and check if there are any leftovers.

Fixes a50245ede72a ("examples/tep_term: initialize VXLAN sample")
---
 examples/tep_termination/Makefile    |  2 +
 examples/tep_termination/main.c      | 70 ++++++++++------------------
 examples/tep_termination/meson.build |  1 +
 3 files changed, 28 insertions(+), 45 deletions(-)
examples/tep_termination/Makefile +2 -0
index 645112498d33..f55331fb24c2 100644
@@ -26,6 +26,7 @@ CFLAGS += -O3 $(shell $(PKGCONF) --cflags libdpdk)
 LDFLAGS_SHARED = $(shell $(PKGCONF) --libs libdpdk)
 LDFLAGS_STATIC = -Wl,-Bstatic $(shell $(PKGCONF) --static --libs libdpdk)
 
+CFLAGS += -DALLOW_EXPERIMENTAL_API
 CFLAGS += -Wno-deprecated-declarations
 
 build/$(APP)-shared: $(SRCS-y) Makefile $(PC_FILE) | build
@@ -61,6 +62,7 @@ endif
 CFLAGS += -O3
 CFLAGS += $(WERROR_FLAGS)
 CFLAGS += -Wno-deprecated-declarations
+CFLAGS += -DALLOW_EXPERIMENTAL_API
 
 include $(RTE_SDK)/mk/rte.extapp.mk
 endif
examples/tep_termination/main.c +25 -45

113 lines not on this page · raw message

examples/tep_termination/meson.build +1 -0
index f65d689802a3..ddadf1c2aede 100644
@@ -9,6 +9,7 @@
 if not is_linux
 	build = false
 endif
+allow_experimental_apis = true
 deps += ['hash', 'vhost']
 cflags += '-Wno-deprecated-declarations'
 sources = files(
Signature · 1 line
2.20.1
#30 lore raw

Re: [dpdk-dev] [dpdk-stable] [PATCH v3 6/6] examples/tep_termination: rework the port setup logic

Quoted 33 lines · Stephen Hemminger

The handling of ports in this application had many problems. It was checking for things that can never happen with current DPDK library (like rte_ethdev_avail_count() >= RTE_MAX_ETHPORTS) and it was not checking if the port was owned and should not be used directly. Also the variable nb_ports was used as both local and global variable.

Fix by rewriting the initialization logic to iterate over valid ports and check if there are any leftovers.

Fixes a50245ede72a ("examples/tep_term: initialize VXLAN sample")
---
 examples/tep_termination/Makefile    |  2 +
 examples/tep_termination/main.c      | 70 ++++++++++------------------
 examples/tep_termination/meson.build |  1 +
 3 files changed, 28 insertions(+), 45 deletions(-)
examples/tep_termination/Makefile +1 -0
index 645112498d33..f55331fb24c2 100644
@@ -26,6 +26,7 @@ CFLAGS += -O3 $(shell $(PKGCONF) --cflags libdpdk)
 LDFLAGS_SHARED = $(shell $(PKGCONF) --libs libdpdk)
 LDFLAGS_STATIC = -Wl,-Bstatic $(shell $(PKGCONF) --static --libs libdpdk)
 
+CFLAGS += -DALLOW_EXPERIMENTAL_API
 CFLAGS += -Wno-deprecated-declarations
 
 build/$(APP)-shared: $(SRCS-y) Makefile $(PC_FILE) | build

examples/tep_termination/Makefile endif · line 65

Quoted 4 lines
examples/tep_termination/Makefile +1 -0
@@ -61,6 +62,7 @@ endif
 CFLAGS += -O3
 CFLAGS += $(WERROR_FLAGS)
 CFLAGS += -Wno-deprecated-declarations
+CFLAGS += -DALLOW_EXPERIMENTAL_API

Why 'ALLOW_EXPERIMENTAL_API' added in this patch, what is the experimental API started to use?

<...>

examples/tep_termination/Makefile endif

Quoted 7 lines
examples/tep_termination/Makefile +0 -6
-	for (portid = 0; portid < nb_ports; portid++) {
-		if (!rte_eth_dev_is_valid_port(ports[portid])) {
-			RTE_LOG(INFO, VHOST_PORT,
-				"\nSpecified port ID(%u) is not valid\n",
-				ports[portid]);
-			ports[portid] = INVALID_PORT_ID;

'INVALID_PORT_ID' macro seems can be removed now.