ethdev: introduce ethdev dump API
RFC 1/1 ethdev: introduce ethdev dump API
Added the ethdev dump API which provides functions for query private info from device. There exists many private properties in different PMD drivers, such as adapter state, Rx/Tx func algorithm in hns3 PMD. The information of these properties is important for debug. As the information is private, the new API is introduced.
--- doc/guides/rel_notes/release_22_03.rst | 6 ++++++ lib/ethdev/ethdev_driver.h | 17 +++++++++++++++++ lib/ethdev/rte_ethdev.c | 15 +++++++++++++++ lib/ethdev/rte_ethdev.h | 16 ++++++++++++++++ 4 files changed, 54 insertions(+)
doc/guides/rel_notes/release_22_03.rst +6 -0
@@ -55,6 +55,12 @@ New Features Also, make sure to start the actual text at the margin. ======================================================= + * **Added the ethdev dump API, for query private info of ethdev.** + + Added the ethdev dump API which provides functions for query private info + from device. There exists many private properties in different PMD + drivers. The information of these properties is important for debug. As + the information is private, the new API is introduced. Removed Items -------------
lib/ethdev/ethdev_driver.h +17 -0
@@ -990,6 +990,20 @@ typedef int (*eth_representor_info_get_t)(struct rte_eth_dev *dev,typedef int (*eth_rx_metadata_negotiate_t )(struct rte_eth_dev *dev,uint64_t *features); +/** +* @internal +* Get ethdev private info. +* +* @param dev +* Port (ethdev) handle. +* @param file +* A pointer to a file for output. +* +* @return +* Negative errno value on error, positive value on success. +*/ +typedef int (*eth_dev_dump_t )(struct rte_eth_dev *dev,FILE *file); +/** * @internal A structure containing the functions exported by an Ethernet driver. */
@@ -1186,6 +1200,9 @@ struct eth_dev_ops {* kinds of metadata to the PMD */ eth_rx_metadata_negotiate_t rx_metadata_negotiate; + +/** Dump ethdev private info */ +eth_dev_dump_t eth_dev_dump; };/**
lib/ethdev/rte_ethdev.c +15 -0
@@ -6472,6 +6472,21 @@ rte_eth_rx_metadata_negotiate(uint16_t port_id, uint64_t *features) (*dev->dev_ops->rx_metadata_negotiate)(dev, features)); } +int +rte_eth_dev_dump (uint16_t port_id,FILE *file) +{ +struct rte_eth_dev *dev; +int ret; + +RTE_ETH_VALID_PORTID_OR_ERR_RET (port_id, -ENODEV ); + dev = &rte_eth_devices[port_id]; + +RTE_FUNC_PTR_OR_ERR_RET (*dev->dev_ops->eth_dev_dump, -ENOTSUP ); + ret = (*dev->dev_ops->eth_dev_dump)(dev, file); + +return ret; +} +RTE_LOG_REGISTER_DEFAULT (rte_eth_dev_logtype,INFO );RTE_INIT (ethdev_init_telemetry)
lib/ethdev/rte_ethdev.h +16 -0
@@ -5888,6 +5888,22 @@ rte_eth_tx_buffer(uint16_t port_id, uint16_t queue_id,return rte_eth_tx_buffer_flush (port_id, queue_id, buffer); } +/** +* @warning +* @b EXPERIMENTAL: this API may change, or be removed, without prior notice +* +* Get ethdev private info. +* +* @param port_id +* The port identifier of the Ethernet device. +* @param file +* A pointer to a file for output. +* @return +* Negative errno value on error, positive value on success. +*/ +__rte_experimental +int rte_eth_dev_dump(uint16_t port_id, FILE *file); + #ifdef __cplusplus }#endif
Signature · 1 line
2.33.0
Quoted 136 lines
From: Min Hu (Connor) [mailto:[email protected]]
Sent: Tuesday, 11 January 2022 12.55Added the ethdev dump API which provides functions for query private info
from device. There exists many private properties in different PMD drivers,
such as adapter state, Rx/Tx func algorithm in hns3 PMD. The information of
these properties is important for debug. As the information is private, the new API is introduced.Signed-off-by Min Hu (Connor) <[email protected]>--- doc/guides/rel_notes/release_22_03.rst | 6 ++++++ lib/ethdev/ethdev_driver.h | 17 +++++++++++++++++ lib/ethdev/rte_ethdev.c | 15 +++++++++++++++ lib/ethdev/rte_ethdev.h | 16 ++++++++++++++++ 4 files changed, 54 insertions(+)doc/guides/rel_notes/release_22_03.rst +3
-0@@ -55,6 +55,12 @@ New Features Also, make sure to start the actual text at the margin. ======================================================= + * **Added the ethdev dump API, for query private info of ethdev.** + + Added the ethdev dump API which provides functions for queryprivate info
+ from device. There exists many private properties in different
PMD
+ drivers. The information of these properties is important for
debug. Asdoc/guides/rel_notes/release_22_03.rst +1
-0+ the information is private, the new API is introduced. Removed Items -------------lib/ethdev/ethdev_driver.h +14
-0@@ -990,6 +990,20 @@ typedef int (*eth_representor_info_get_t)(structrte_eth_dev *dev, typedef int (*eth_rx_metadata_negotiate_t)(struct rte_eth_dev *dev,uint64_t *features); +/** +* @internal +* Get ethdev private info. +* +* @param dev +* Port (ethdev) handle. +* @param file +* A pointer to a file for output. +* +* @return +* Negative errno value on error, positive value on success. +*/ +typedef int (*eth_dev_dump_t )(struct rte_eth_dev *dev,FILE *file); +/** * @internal A structure containing the functions exported by an Ethernet driver.
*/lib/ethdev/ethdev_driver.h +3
-0@@ -1186,6 +1200,9 @@ struct eth_dev_ops {* kinds of metadata to the PMD */ eth_rx_metadata_negotiate_t rx_metadata_negotiate; + +/** Dump ethdev private info */ +eth_dev_dump_t eth_dev_dump; };/** lib/ethdev/rte_ethdev.c +15
-0@@ -6472,6 +6472,21 @@ rte_eth_rx_metadata_negotiate(uint16_t port_id, uint64_t *features) (*dev->dev_ops->rx_metadata_negotiate)(dev, features)); } +int +rte_eth_dev_dump(uint16_t port_id, FILE *file) +{ + struct rte_eth_dev *dev; + int ret; + + RTE_ETH_VALID_PORTID_OR_ERR_RET(port_id, -ENODEV); + dev = &rte_eth_devices[port_id]; + + RTE_FUNC_PTR_OR_ERR_RET(*dev->dev_ops->eth_dev_dump, -ENOTSUP); + ret = (*dev->dev_ops->eth_dev_dump)(dev, file); + + return ret; +} + RTE_LOG_REGISTER_DEFAULT(rte_eth_dev_logtype, INFO);
RTE_INIT(ethdev_init_telemetry)lib/ethdev/rte_ethdev.h +3
-0@@ -5888,6 +5888,22 @@ rte_eth_tx_buffer(uint16_t port_id, uint16_t queue_id, return rte_eth_tx_buffer_flush(port_id, queue_id, buffer); } +/** +* @warning +* @b EXPERIMENTAL: this API may change, or be removed, without prior notice
+ *
+ * Get ethdev private info.
+ *
+ * @paramport_id
+ * The port identifier of the Ethernet device.
+ * @param file
+ * A pointer to a file for output.
+ * @return
+ * Negative errno value on error, positive value on success.
+ */lib/ethdev/rte_ethdev.h +3
-1+__rte_experimental +int rte_eth_dev_dump(uint16_t port_id, FILE *file); + #ifdef __cplusplus }#endif --
Quoted 2 lines
2.33.0
Good idea.
Two comments:
1. This function dumps private ethdev information. It should be named and described as such, e.g. eth_dev_priv_dump().
It should be a generic ethdev dump function, which dumps the common ethdev information, and also calls the eth_dev_priv_dump() function, if present.
2. Please make FILE* the first parameter, like similar dump functions.
Hi, Morten,
thanks for your reply, all is fixed in v2.
Quoted 152 lines · Morten Brørup
Quoted 136 lines
From: Min Hu (Connor) [mailto:[email protected]]
Sent: Tuesday, 11 January 2022 12.55Added the ethdev dump API which provides functions for query private info
from device. There exists many private properties in different PMD drivers,
such as adapter state, Rx/Tx func algorithm in hns3 PMD. The information of
these properties is important for debug. As the information is private, the new API is introduced.Signed-off-by Min Hu (Connor) <[email protected]>--- doc/guides/rel_notes/release_22_03.rst | 6 ++++++ lib/ethdev/ethdev_driver.h | 17 +++++++++++++++++ lib/ethdev/rte_ethdev.c | 15 +++++++++++++++ lib/ethdev/rte_ethdev.h | 16 ++++++++++++++++ 4 files changed, 54 insertions(+)doc/guides/rel_notes/release_22_03.rst +3
-0@@ -55,6 +55,12 @@ New Features Also, make sure to start the actual text at the margin. ======================================================= + * **Added the ethdev dump API, for query private info of ethdev.** + + Added the ethdev dump API which provides functions for queryprivate info
+ from device. There exists many private properties in different
PMD
+ drivers. The information of these properties is important for
debug. Asdoc/guides/rel_notes/release_22_03.rst +1
-0+ the information is private, the new API is introduced. Removed Items -------------lib/ethdev/ethdev_driver.h +14
-0@@ -990,6 +990,20 @@ typedef int (*eth_representor_info_get_t)(structrte_eth_dev *dev, typedef int (*eth_rx_metadata_negotiate_t)(struct rte_eth_dev *dev,uint64_t *features); +/** +* @internal +* Get ethdev private info. +* +* @param dev +* Port (ethdev) handle. +* @param file +* A pointer to a file for output. +* +* @return +* Negative errno value on error, positive value on success. +*/ +typedef int (*eth_dev_dump_t )(struct rte_eth_dev *dev,FILE *file); +/** * @internal A structure containing the functions exported by an Ethernet driver.
*/lib/ethdev/ethdev_driver.h +3
-0@@ -1186,6 +1200,9 @@ struct eth_dev_ops {* kinds of metadata to the PMD */ eth_rx_metadata_negotiate_t rx_metadata_negotiate; + +/** Dump ethdev private info */ +eth_dev_dump_t eth_dev_dump; };/** lib/ethdev/rte_ethdev.c +15
-0@@ -6472,6 +6472,21 @@ rte_eth_rx_metadata_negotiate(uint16_t port_id, uint64_t *features) (*dev->dev_ops->rx_metadata_negotiate)(dev, features)); } +int +rte_eth_dev_dump(uint16_t port_id, FILE *file) +{ + struct rte_eth_dev *dev; + int ret; + + RTE_ETH_VALID_PORTID_OR_ERR_RET(port_id, -ENODEV); + dev = &rte_eth_devices[port_id]; + + RTE_FUNC_PTR_OR_ERR_RET(*dev->dev_ops->eth_dev_dump, -ENOTSUP); + ret = (*dev->dev_ops->eth_dev_dump)(dev, file); + + return ret; +} + RTE_LOG_REGISTER_DEFAULT(rte_eth_dev_logtype, INFO);
RTE_INIT(ethdev_init_telemetry)lib/ethdev/rte_ethdev.h +3
-0@@ -5888,6 +5888,22 @@ rte_eth_tx_buffer(uint16_t port_id, uint16_t queue_id, return rte_eth_tx_buffer_flush(port_id, queue_id, buffer); } +/** +* @warning +* @b EXPERIMENTAL: this API may change, or be removed, without prior notice
+ *
+ * Get ethdev private info.
+ *
+ * @paramport_id
+ * The port identifier of the Ethernet device.
+ * @param file
+ * A pointer to a file for output.
+ * @return
+ * Negative errno value on error, positive value on success.
+ */lib/ethdev/rte_ethdev.h +3
-1+__rte_experimental +int rte_eth_dev_dump(uint16_t port_id, FILE *file); + #ifdef __cplusplus }#endif --Quoted 2 lines
2.33.0
Good idea.
Two comments:
1. This function dumps private ethdev information. It should be named and described as such, e.g.
eth_dev_priv_dump().It should be a generic ethdev dump function, which dumps the common ethdev information, and also calls the
eth_dev_priv_dump()function, if present.2. Please make FILE* the first parameter, like similar dump functions.
.
Please use --cc-cmd devtools/get-maintainer.sh so all maintainers are Cc'ed.
11/01/2022 12:54, Min Hu (Connor):
doc/guides/rel_notes/release_22_03.rst New Features · line 64
Quoted 13 lines
doc/guides/rel_notes/release_22_03.rst +6
-0@@ -55,6 +55,12 @@ New Features Also, make sure to start the actual text at the margin. ======================================================= + * **Added the ethdev dump API, for query private info of ethdev.** + + Added the ethdev dump API which provides functions for query private info + from device. There exists many private properties in different PMD + drivers. The information of these properties is important for debug. As + the information is private, the new API is introduced.
A blank line is missing.
Also please check the comment above asking to start the actual text at the margin.
[...]
doc/guides/rel_notes/release_22_03.rst New Features · line 65
Quoted 1 line
doc/guides/rel_notes/release_22_03.rst +1
-0+typedef int (*eth_dev_dump_t)(struct rte_eth_dev *dev, FILE *file);
There is a dump function for rte_flow: rte_flow_dev_dump(). This one should have a clear scope: private device infos?
Hi, Thomas,
fixed in v2, thanks.
Quoted 30 lines · Thomas Monjalon
Please use --cc-cmd
devtools/get-maintainer.shso all maintainers are Cc'ed.11/01/2022 12:54, Min Hu (Connor):
doc/guides/rel_notes/release_22_03.rst New Features · line 64
Quoted 13 lines
doc/guides/rel_notes/release_22_03.rst +6
-0@@ -55,6 +55,12 @@ New Features Also, make sure to start the actual text at the margin. ======================================================= + * **Added the ethdev dump API, for query private info of ethdev.** + + Added the ethdev dump API which provides functions for query private info + from device. There exists many private properties in different PMD + drivers. The information of these properties is important for debug. As + the information is private, the new API is introduced.A blank line is missing.
Also please check the comment above asking to start the actual text at the margin.[...]
doc/guides/rel_notes/release_22_03.rst New Features · line 65
Quoted 1 line
doc/guides/rel_notes/release_22_03.rst +1
-0+typedef int (*eth_dev_dump_t)(struct rte_eth_dev *dev, FILE *file);There is a dump function for
rte_flow:rte_flow_dev_dump(). This one should have a clear scope: private device infos?.
Hi, Thomas,
Quoted 1 line · Thomas Monjalon
Please use --cc-cmd
devtools/get-maintainer.shso all maintainers are Cc'edlike this ?
git send-email -to [email protected] -cc [email protected] -cc
[email protected] --cc-cmd devtools/get-maintainer.sh *.patch --suppress-cc=all
I did this, but it doesn't work. It only Cc ferruh and you.
Quoted 29 lines
11/01/2022 12:54, Min Hu (Connor):
doc/guides/rel_notes/release_22_03.rst New Features · line 64
Quoted 13 lines
doc/guides/rel_notes/release_22_03.rst +6
-0@@ -55,6 +55,12 @@ New Features Also, make sure to start the actual text at the margin. ======================================================= + * **Added the ethdev dump API, for query private info of ethdev.** + + Added the ethdev dump API which provides functions for query private info + from device. There exists many private properties in different PMD + drivers. The information of these properties is important for debug. As + the information is private, the new API is introduced.A blank line is missing.
Also please check the comment above asking to start the actual text at the margin.[...]
doc/guides/rel_notes/release_22_03.rst New Features · line 65
Quoted 1 line
doc/guides/rel_notes/release_22_03.rst +1
-0+typedef int (*eth_dev_dump_t)(struct rte_eth_dev *dev, FILE *file);There is a dump function for
rte_flow:rte_flow_dev_dump(). This one should have a clear scope: private device infos?.
12/01/2022 03:44, Min Hu (Connor):
Quoted 9 lines
Hi, Thomas,
Quoted 1 line · Thomas Monjalon
Please use --cc-cmd
devtools/get-maintainer.shso all maintainers are Cc'edlike this ?git send-email -to [email protected] -cc [email protected] -cc
[email protected] --cc-cmddevtools/get-maintainer.sh*.patch --suppress-cc=allI did this, but it doesn't work. It only Cc ferruh and you.
You should read the documentation about the tools:
https://doc.dpdk.org/guides/contributing/patches.html#checking-the-patches
In my file ~/.config/dpdk/devel.config, I have these lines:
export DPDK_GETMAINTAINER_PATH=$root/linux/scripts/get_maintainer.pl export DPDK_CHECKPATCH_PATH=$root/linux/scripts/checkpatch.pl export DPDK_CHECKPATCH_CODESPELL=$root/codespell/dictionary.txt
Did you set DPDK_GETMAINTAINER_PATH?
Thanks Thomas, I will have a try.
Quoted 24 lines · Thomas Monjalon
12/01/2022 03:44, Min Hu (Connor):
Quoted 9 lines
Hi, Thomas,
Quoted 1 line · Thomas Monjalon
Please use --cc-cmd
devtools/get-maintainer.shso all maintainers are Cc'edlike this ?git send-email -to [email protected] -cc [email protected] -cc
[email protected] --cc-cmddevtools/get-maintainer.sh*.patch --suppress-cc=allI did this, but it doesn't work. It only Cc ferruh and you.
You should read the documentation about the tools:
https://doc.dpdk.org/guides/contributing/patches.html#checking-the-patchesIn my file
~/.config/dpdk/devel.config, I have these lines:
export DPDK_GETMAINTAINER_PATH=$root/linux/scripts/get_maintainer.pl export DPDK_CHECKPATCH_PATH=$root/linux/scripts/checkpatch.pl export DPDK_CHECKPATCH_CODESPELL=$root/codespell/dictionary.txtDid you set
DPDK_GETMAINTAINER_PATH?.
RFC v2 1/1 ethdev: introduce ethdev dump API
In reply to Min Hu (Connor)Added the ethdev dump API which provides functions for query private info from device. There exists many private properties in different PMD drivers, such as adapter state, Rx/Tx func algorithm in hns3 PMD. The information of these properties is important for debug. As the information is private, the new API is introduced.
---
v2:
* fix dump API name
* adjust description in doc.
--- doc/guides/rel_notes/release_22_03.rst | 7 +++++++ lib/ethdev/ethdev_driver.h | 17 +++++++++++++++++ lib/ethdev/rte_ethdev.c | 15 +++++++++++++++ lib/ethdev/rte_ethdev.h | 16 ++++++++++++++++ 4 files changed, 55 insertions(+)
doc/guides/rel_notes/release_22_03.rst +7 -0
@@ -55,6 +55,13 @@ New Features Also, make sure to start the actual text at the margin. ======================================================= +* **Added the private ethdev dump API, for query private info of ethdev.** + + Added the private ethdev dump API which provides functions for query + private info from device. There exists many private properties in + different PMD drivers. The information of these properties is important + for debug. As the information is private, the new API is introduced. + Removed Items -------------
lib/ethdev/ethdev_driver.h +17 -0
@@ -990,6 +990,20 @@ typedef int (*eth_representor_info_get_t)(struct rte_eth_dev *dev,typedef int (*eth_rx_metadata_negotiate_t )(struct rte_eth_dev *dev,uint64_t *features); +/** +* @internal +* Get ethdev private info. +* +* @param file +* A pointer to a file for output. +* @param dev +* Port (ethdev) handle. +* +* @return +* Negative errno value on error, positive value on success. +*/ +typedef int (*eth_dev_priv_dump_t )(FILE *file,struct rte_eth_dev *dev); +/** * @internal A structure containing the functions exported by an Ethernet driver. */
@@ -1186,6 +1200,9 @@ struct eth_dev_ops {* kinds of metadata to the PMD */ eth_rx_metadata_negotiate_t rx_metadata_negotiate; + +/** Dump ethdev private info */ +eth_dev_priv_dump_t eth_dev_priv_dump; };/**
lib/ethdev/rte_ethdev.c +15 -0
@@ -6472,6 +6472,21 @@ rte_eth_rx_metadata_negotiate(uint16_t port_id, uint64_t *features) (*dev->dev_ops->rx_metadata_negotiate)(dev, features)); } +int +rte_eth_dev_priv_dump (FILE *file,uint16_t port_id) +{ +struct rte_eth_dev *dev; +int ret; + +RTE_ETH_VALID_PORTID_OR_ERR_RET (port_id, -ENODEV ); + dev = &rte_eth_devices[port_id]; + +RTE_FUNC_PTR_OR_ERR_RET (*dev->dev_ops->eth_dev_priv_dump, -ENOTSUP ); + ret = (*dev->dev_ops->eth_dev_priv_dump)(file, dev); + +return ret; +} +RTE_LOG_REGISTER_DEFAULT (rte_eth_dev_logtype,INFO );RTE_INIT (ethdev_init_telemetry)
lib/ethdev/rte_ethdev.h +16 -0
@@ -5888,6 +5888,22 @@ rte_eth_tx_buffer(uint16_t port_id, uint16_t queue_id,return rte_eth_tx_buffer_flush (port_id, queue_id, buffer); } +/** +* @warning +* @b EXPERIMENTAL: this API may change, or be removed, without prior notice +* +* Get ethdev private info. +* +* @param file +* A pointer to a file for output. +* @param port_id +* The port identifier of the Ethernet device. +* @return +* Negative errno value on error, positive value on success. +*/ +__rte_experimental +int rte_eth_dev_priv_dump(FILE *file, uint16_t port_id); + #ifdef __cplusplus }#endif
Signature · 1 line
2.33.0
Quoted 50 lines
From: Min Hu (Connor) [mailto:[email protected]]
Sent: Wednesday, 12 January 2022 03.40Added the ethdev dump API which provides functions for query private info
from device. There exists many private properties in different PMD drivers,
such as adapter state, Rx/Tx func algorithm in hns3 PMD. The information of
these properties is important for debug. As the information is private, the new API is introduced.Signed-off-by Min Hu (Connor) <[email protected]>---
v2:
* fix dump API name
* adjust description in doc.--- doc/guides/rel_notes/release_22_03.rst | 7 +++++++ lib/ethdev/ethdev_driver.h | 17 +++++++++++++++++ lib/ethdev/rte_ethdev.c | 15 +++++++++++++++ lib/ethdev/rte_ethdev.h | 16 ++++++++++++++++ 4 files changed, 55 insertions(+)doc/guides/rel_notes/release_22_03.rst +1
-0@@ -55,6 +55,13 @@ New Features Also, make sure to start the actual text at the margin. ======================================================= +* **Added the private ethdev dump API, for query private info ofethdev.**
doc/guides/rel_notes/release_22_03.rst +4
-0+ + Added the private ethdev dump API which provides functions for query + private info from device. There exists many private properties in + different PMD drivers. The information of these properties isimportant
doc/guides/rel_notes/release_22_03.rst +2
-0+ for debug. As the information is private, the new API is introduced. + Removed Items -------------lib/ethdev/ethdev_driver.h +0
-0
lib/ethdev/ethdev_driver.h typedef int (*eth_representor_info_get_t)(struct · line 996
Quoted 7 lines
lib/ethdev/ethdev_driver.h +3
-0@@ -990,6 +990,20 @@ typedef int (*eth_representor_info_get_t)(structrte_eth_dev *dev, typedef int (*eth_rx_metadata_negotiate_t)(struct rte_eth_dev *dev,uint64_t *features); +/** +* @internal +* Get ethdev private info.
Suggestion: Dump ethdev private info to a file.
Quoted 67 lines
lib/ethdev/ethdev_driver.h +11
-0+* +* @param file +* A pointer to a file for output. +* @param dev +* Port (ethdev) handle. +* +* @return +* Negative errno value on error, positive value on success. +*/ +typedef int (*eth_dev_priv_dump_t )(FILE *file,struct rte_eth_dev *dev); +/** * @internal A structure containing the functions exported by an
Ethernet driver.
*/lib/ethdev/ethdev_driver.h +3
-0@@ -1186,6 +1200,9 @@ struct eth_dev_ops {* kinds of metadata to the PMD */ eth_rx_metadata_negotiate_t rx_metadata_negotiate; + +/** Dump ethdev private info */ +eth_dev_priv_dump_t eth_dev_priv_dump; };/** lib/ethdev/rte_ethdev.c +15
-0@@ -6472,6 +6472,21 @@ rte_eth_rx_metadata_negotiate(uint16_t port_id, uint64_t *features) (*dev->dev_ops->rx_metadata_negotiate)(dev, features)); } +int +rte_eth_dev_priv_dump(FILE *file, uint16_t port_id) +{ + struct rte_eth_dev *dev; + int ret; + + RTE_ETH_VALID_PORTID_OR_ERR_RET(port_id, -ENODEV); + dev = &rte_eth_devices[port_id]; + + RTE_FUNC_PTR_OR_ERR_RET(*dev->dev_ops->eth_dev_priv_dump, - ENOTSUP); + ret = (*dev->dev_ops->eth_dev_priv_dump)(file, dev); + + return ret; +} +
RTE_LOG_REGISTER_DEFAULT(rte_eth_dev_logtype, INFO);
RTE_INIT(ethdev_init_telemetry)lib/ethdev/rte_ethdev.h +3
-0@@ -5888,6 +5888,22 @@ rte_eth_tx_buffer(uint16_t port_id, uint16_t queue_id, return rte_eth_tx_buffer_flush(port_id, queue_id, buffer); } +/** +* @warning +* @b EXPERIMENTAL: this API may change, or be removed, without prior
Quoted 3 lines
notice
+ *
+ * Get ethdev private info.
Suggestion: Dump ethdev private info to a file.
Quoted 15 lines
+ *
+ * @param file
+ * A pointer to a file for output.
+ * @paramport_id
+ * The port identifier of the Ethernet device.
+ * @return
+ * Negative errno value on error, positive value on success.
+ */lib/ethdev/rte_ethdev.h +3
-1+__rte_experimental +int rte_eth_dev_priv_dump(FILE *file, uint16_t port_id); + #ifdef __cplusplus }#endif --
Quoted 2 lines
2.33.0
You should probably also add rte_eth_dev_priv_dump to the EXPERIMENTAL section of the version.map file.
Thanks Morten, fixed in v3.
Quoted 157 lines · Morten Brørup
Quoted 50 lines
From: Min Hu (Connor) [mailto:[email protected]]
Sent: Wednesday, 12 January 2022 03.40Added the ethdev dump API which provides functions for query private info
from device. There exists many private properties in different PMD drivers,
such as adapter state, Rx/Tx func algorithm in hns3 PMD. The information of
these properties is important for debug. As the information is private, the new API is introduced.Signed-off-by Min Hu (Connor) <[email protected]>---
v2:
* fix dump API name
* adjust description in doc.--- doc/guides/rel_notes/release_22_03.rst | 7 +++++++ lib/ethdev/ethdev_driver.h | 17 +++++++++++++++++ lib/ethdev/rte_ethdev.c | 15 +++++++++++++++ lib/ethdev/rte_ethdev.h | 16 ++++++++++++++++ 4 files changed, 55 insertions(+)doc/guides/rel_notes/release_22_03.rst +1
-0@@ -55,6 +55,13 @@ New Features Also, make sure to start the actual text at the margin. ======================================================= +* **Added the private ethdev dump API, for query private info ofethdev.**
doc/guides/rel_notes/release_22_03.rst +4
-0+ + Added the private ethdev dump API which provides functions for query + private info from device. There exists many private properties in + different PMD drivers. The information of these properties isimportant
doc/guides/rel_notes/release_22_03.rst +2
-0+ for debug. As the information is private, the new API is introduced. + Removed Items -------------lib/ethdev/ethdev_driver.h +0
-0lib/ethdev/ethdev_driver.h typedef int (*eth_representor_info_get_t)(struct · line 996
Quoted 7 lines
lib/ethdev/ethdev_driver.h +3
-0@@ -990,6 +990,20 @@ typedef int (*eth_representor_info_get_t)(structrte_eth_dev *dev, typedef int (*eth_rx_metadata_negotiate_t)(struct rte_eth_dev *dev,uint64_t *features); +/** +* @internal +* Get ethdev private info. Suggestion: Dump ethdev private info to a file.
Quoted 67 lines
lib/ethdev/ethdev_driver.h +11
-0+* +* @param file +* A pointer to a file for output. +* @param dev +* Port (ethdev) handle. +* +* @return +* Negative errno value on error, positive value on success. +*/ +typedef int (*eth_dev_priv_dump_t )(FILE *file,struct rte_eth_dev *dev); +/** * @internal A structure containing the functions exported by an
Ethernet driver.
*/lib/ethdev/ethdev_driver.h +3
-0@@ -1186,6 +1200,9 @@ struct eth_dev_ops {* kinds of metadata to the PMD */ eth_rx_metadata_negotiate_t rx_metadata_negotiate; + +/** Dump ethdev private info */ +eth_dev_priv_dump_t eth_dev_priv_dump; };/** lib/ethdev/rte_ethdev.c +15
-0@@ -6472,6 +6472,21 @@ rte_eth_rx_metadata_negotiate(uint16_t port_id, uint64_t *features) (*dev->dev_ops->rx_metadata_negotiate)(dev, features)); } +int +rte_eth_dev_priv_dump(FILE *file, uint16_t port_id) +{ + struct rte_eth_dev *dev; + int ret; + + RTE_ETH_VALID_PORTID_OR_ERR_RET(port_id, -ENODEV); + dev = &rte_eth_devices[port_id]; + + RTE_FUNC_PTR_OR_ERR_RET(*dev->dev_ops->eth_dev_priv_dump, - ENOTSUP); + ret = (*dev->dev_ops->eth_dev_priv_dump)(file, dev); + + return ret; +} +RTE_LOG_REGISTER_DEFAULT(rte_eth_dev_logtype, INFO);
RTE_INIT(ethdev_init_telemetry)lib/ethdev/rte_ethdev.h +3
-0@@ -5888,6 +5888,22 @@ rte_eth_tx_buffer(uint16_t port_id, uint16_t queue_id, return rte_eth_tx_buffer_flush(port_id, queue_id, buffer); } +/** +* @warning +* @b EXPERIMENTAL: this API may change, or be removed, without prior Quoted 3 lines
notice
+ *
+ * Get ethdev private info.Suggestion: Dump ethdev private info to a file.
Quoted 15 lines
+ *
+ * @param file
+ * A pointer to a file for output.
+ * @paramport_id
+ * The port identifier of the Ethernet device.
+ * @return
+ * Negative errno value on error, positive value on success.
+ */lib/ethdev/rte_ethdev.h +3
-1+__rte_experimental +int rte_eth_dev_priv_dump(FILE *file, uint16_t port_id); + #ifdef __cplusplus }#endif --Quoted 2 lines
2.33.0
You should probably also add
rte_eth_dev_priv_dumpto the EXPERIMENTAL section of the version.map file.Acked-by Morten Brørup <[email protected]>.
Quoted 15 lines · Min Hu (Connor)
Thanks Morten, fixed in v3.
Quoted 11 lines · Morten Brørup
Quoted 11 lines
From: Min Hu (Connor) [mailto:[email protected]]
Sent: Wednesday, 12 January 2022 03.40Added the ethdev dump API which provides functions for query private info
from device. There exists many private properties in different PMD drivers,
such as adapter state, Rx/Tx func algorithm in hns3 PMD. The information of
these properties is important for debug. As the information is private, the new API is introduced.
Do you have any changes to testpmd to use this API?
Quoted 146 lines
Quoted 146 lines
Quoted 39 lines
Signed-off-by Min Hu (Connor) <[email protected]>---
v2:
* fix dump API name
* adjust description in doc.--- doc/guides/rel_notes/release_22_03.rst | 7 +++++++ lib/ethdev/ethdev_driver.h | 17 +++++++++++++++++ lib/ethdev/rte_ethdev.c | 15 +++++++++++++++ lib/ethdev/rte_ethdev.h | 16 ++++++++++++++++ 4 files changed, 55 insertions(+)doc/guides/rel_notes/release_22_03.rst +1
-0@@ -55,6 +55,13 @@ New Features Also, make sure to start the actual text at the margin. ======================================================= +* **Added the private ethdev dump API, for query private info ofethdev.**
doc/guides/rel_notes/release_22_03.rst +4
-0+ + Added the private ethdev dump API which provides functions for query + private info from device. There exists many private properties in + different PMD drivers. The information of these properties isimportant
doc/guides/rel_notes/release_22_03.rst +2
-0+ for debug. As the information is private, the new API is introduced. + Removed Items -------------lib/ethdev/ethdev_driver.h +0
-0lib/ethdev/ethdev_driver.h typedef int (*eth_representor_info_get_t)(struct · line 996
Quoted 7 lines
lib/ethdev/ethdev_driver.h +3
-0@@ -990,6 +990,20 @@ typedef int (*eth_representor_info_get_t)(structrte_eth_dev *dev, typedef int (*eth_rx_metadata_negotiate_t)(struct rte_eth_dev *dev,uint64_t *features); +/** +* @internal +* Get ethdev private info. Suggestion: Dump ethdev private info to a file.
Quoted 67 lines
lib/ethdev/ethdev_driver.h +11
-0+* +* @param file +* A pointer to a file for output. +* @param dev +* Port (ethdev) handle. +* +* @return +* Negative errno value on error, positive value on success. +*/ +typedef int (*eth_dev_priv_dump_t )(FILE *file,struct rte_eth_dev *dev); +/** * @internal A structure containing the functions exported by an
Ethernet driver.
*/lib/ethdev/ethdev_driver.h +3
-0@@ -1186,6 +1200,9 @@ struct eth_dev_ops {* kinds of metadata to the PMD */ eth_rx_metadata_negotiate_t rx_metadata_negotiate; + +/** Dump ethdev private info */ +eth_dev_priv_dump_t eth_dev_priv_dump; };/** lib/ethdev/rte_ethdev.c +15
-0@@ -6472,6 +6472,21 @@ rte_eth_rx_metadata_negotiate(uint16_t port_id, uint64_t *features) (*dev->dev_ops->rx_metadata_negotiate)(dev, features)); } +int +rte_eth_dev_priv_dump(FILE *file, uint16_t port_id) +{ + struct rte_eth_dev *dev; + int ret; + + RTE_ETH_VALID_PORTID_OR_ERR_RET(port_id, -ENODEV); + dev = &rte_eth_devices[port_id]; + + RTE_FUNC_PTR_OR_ERR_RET(*dev->dev_ops->eth_dev_priv_dump, - ENOTSUP); + ret = (*dev->dev_ops->eth_dev_priv_dump)(file, dev); + + return ret; +} +RTE_LOG_REGISTER_DEFAULT(rte_eth_dev_logtype, INFO);
RTE_INIT(ethdev_init_telemetry)lib/ethdev/rte_ethdev.h +3
-0@@ -5888,6 +5888,22 @@ rte_eth_tx_buffer(uint16_t port_id, uint16_t queue_id, return rte_eth_tx_buffer_flush(port_id, queue_id, buffer); } +/** +* @warning +* @b EXPERIMENTAL: this API may change, or be removed, without prior Quoted 3 lines
notice
+ *
+ * Get ethdev private info.Suggestion: Dump ethdev private info to a file.
Quoted 15 lines
+ *
+ * @param file
+ * A pointer to a file for output.
+ * @paramport_id
+ * The port identifier of the Ethernet device.
+ * @return
+ * Negative errno value on error, positive value on success.
+ */lib/ethdev/rte_ethdev.h +3
-1+__rte_experimental +int rte_eth_dev_priv_dump(FILE *file, uint16_t port_id); + #ifdef __cplusplus }#endif --Quoted 2 lines
2.33.0
You should probably also add
rte_eth_dev_priv_dumpto the EXPERIMENTAL section of the version.map file.Acked-by Morten Brørup <[email protected]>.
Hi, Ajit,
Quoted 16 lines · Ajit Khaparde
Quoted 15 lines · Min Hu (Connor)
Thanks Morten, fixed in v3.
Quoted 11 lines · Morten Brørup
Quoted 11 lines
From: Min Hu (Connor) [mailto:[email protected]]
Sent: Wednesday, 12 January 2022 03.40Added the ethdev dump API which provides functions for query private info
from device. There exists many private properties in different PMD drivers,
such as adapter state, Rx/Tx func algorithm in hns3 PMD. The information of
these properties is important for debug. As the information is private, the new API is introduced.
Quoted 2 lines
Do you have any changes to testpmd to use this API?
No changes to testpmd.
BTW, I will add some options for proc-info to use this API, and this is my next-step plan.
Quoted 148 lines
Quoted 142 lines
Quoted 142 lines
Quoted 39 lines
Signed-off-by Min Hu (Connor) <[email protected]>---
v2:
* fix dump API name
* adjust description in doc.--- doc/guides/rel_notes/release_22_03.rst | 7 +++++++ lib/ethdev/ethdev_driver.h | 17 +++++++++++++++++ lib/ethdev/rte_ethdev.c | 15 +++++++++++++++ lib/ethdev/rte_ethdev.h | 16 ++++++++++++++++ 4 files changed, 55 insertions(+)doc/guides/rel_notes/release_22_03.rst +1
-0@@ -55,6 +55,13 @@ New Features Also, make sure to start the actual text at the margin. ======================================================= +* **Added the private ethdev dump API, for query private info ofethdev.**
doc/guides/rel_notes/release_22_03.rst +4
-0+ + Added the private ethdev dump API which provides functions for query + private info from device. There exists many private properties in + different PMD drivers. The information of these properties isimportant
doc/guides/rel_notes/release_22_03.rst +2
-0+ for debug. As the information is private, the new API is introduced. + Removed Items -------------lib/ethdev/ethdev_driver.h +0
-0lib/ethdev/ethdev_driver.h typedef int (*eth_representor_info_get_t)(struct · line 996
Quoted 7 lines
lib/ethdev/ethdev_driver.h +3
-0@@ -990,6 +990,20 @@ typedef int (*eth_representor_info_get_t)(structrte_eth_dev *dev, typedef int (*eth_rx_metadata_negotiate_t)(struct rte_eth_dev *dev,uint64_t *features); +/** +* @internal +* Get ethdev private info. Suggestion: Dump ethdev private info to a file.
Quoted 67 lines
lib/ethdev/ethdev_driver.h +11
-0+* +* @param file +* A pointer to a file for output. +* @param dev +* Port (ethdev) handle. +* +* @return +* Negative errno value on error, positive value on success. +*/ +typedef int (*eth_dev_priv_dump_t )(FILE *file,struct rte_eth_dev *dev); +/** * @internal A structure containing the functions exported by an
Ethernet driver.*/lib/ethdev/ethdev_driver.h +3
-0@@ -1186,6 +1200,9 @@ struct eth_dev_ops {* kinds of metadata to the PMD */ eth_rx_metadata_negotiate_t rx_metadata_negotiate; + +/** Dump ethdev private info */ +eth_dev_priv_dump_t eth_dev_priv_dump; };/** lib/ethdev/rte_ethdev.c +15
-0@@ -6472,6 +6472,21 @@ rte_eth_rx_metadata_negotiate(uint16_t port_id, uint64_t *features) (*dev->dev_ops->rx_metadata_negotiate)(dev, features)); } +int +rte_eth_dev_priv_dump(FILE *file, uint16_t port_id) +{ + struct rte_eth_dev *dev; + int ret; + + RTE_ETH_VALID_PORTID_OR_ERR_RET(port_id, -ENODEV); + dev = &rte_eth_devices[port_id]; + + RTE_FUNC_PTR_OR_ERR_RET(*dev->dev_ops->eth_dev_priv_dump, - ENOTSUP); + ret = (*dev->dev_ops->eth_dev_priv_dump)(file, dev); + + return ret; +} +RTE_LOG_REGISTER_DEFAULT(rte_eth_dev_logtype, INFO);
RTE_INIT(ethdev_init_telemetry)lib/ethdev/rte_ethdev.h +3
-0@@ -5888,6 +5888,22 @@ rte_eth_tx_buffer(uint16_t port_id, uint16_t queue_id, return rte_eth_tx_buffer_flush(port_id, queue_id, buffer); } +/** +* @warning +* @b EXPERIMENTAL: this API may change, or be removed, without prior Quoted 3 lines
notice
+ *
+ * Get ethdev private info.Suggestion: Dump ethdev private info to a file.
Quoted 15 lines
+ *
+ * @param file
+ * A pointer to a file for output.
+ * @paramport_id
+ * The port identifier of the Ethernet device.
+ * @return
+ * Negative errno value on error, positive value on success.
+ */lib/ethdev/rte_ethdev.h +3
-1+__rte_experimental +int rte_eth_dev_priv_dump(FILE *file, uint16_t port_id); + #ifdef __cplusplus }#endif --Quoted 2 lines
2.33.0
You should probably also add
rte_eth_dev_priv_dumpto the EXPERIMENTAL section of the version.map file.Quoted 4 lines
Quoted 4 lines
Acked-by Morten Brørup <[email protected]>.
.
Quoted 20 lines · Min Hu (Connor)
Hi, Ajit,
Quoted 16 lines · Ajit Khaparde
Quoted 15 lines · Min Hu (Connor)
Thanks Morten, fixed in v3.
Quoted 11 lines · Morten Brørup
Quoted 11 lines
From: Min Hu (Connor) [mailto:[email protected]]
Sent: Wednesday, 12 January 2022 03.40Added the ethdev dump API which provides functions for query private info
from device. There exists many private properties in different PMD drivers,
such as adapter state, Rx/Tx func algorithm in hns3 PMD. The information of
these properties is important for debug. As the information is private, the new API is introduced.
Quoted 5 lines
Quoted 2 lines
Do you have any changes to testpmd to use this API?
No changes to testpmd.
BTW, I will add some options for proc-info to use this API, and this is my next-step plan.
Ok. proc-info is a good place for this.
Quoted 149 lines
Quoted 149 lines
Quoted 143 lines
Quoted 143 lines
Quoted 40 lines
Signed-off-by Min Hu (Connor) <[email protected]>---
v2:
* fix dump API name
* adjust description in doc.--- doc/guides/rel_notes/release_22_03.rst | 7 +++++++ lib/ethdev/ethdev_driver.h | 17 +++++++++++++++++ lib/ethdev/rte_ethdev.c | 15 +++++++++++++++ lib/ethdev/rte_ethdev.h | 16 ++++++++++++++++ 4 files changed, 55 insertions(+)doc/guides/rel_notes/release_22_03.rst +1
-0@@ -55,6 +55,13 @@ New Features Also, make sure to start the actual text at the margin. ======================================================= +* **Added the private ethdev dump API, for query private info ofethdev.**
doc/guides/rel_notes/release_22_03.rst +4
-0+ + Added the private ethdev dump API which provides functions for query + private info from device. There exists many private properties in + different PMD drivers. The information of these properties isimportant
doc/guides/rel_notes/release_22_03.rst +2
-0+ for debug. As the information is private, the new API is introduced. + Removed Items -------------lib/ethdev/ethdev_driver.h +0
-0lib/ethdev/ethdev_driver.h typedef int (*eth_representor_info_get_t)(struct · line 996
Quoted 7 lines
lib/ethdev/ethdev_driver.h +3
-0@@ -990,6 +990,20 @@ typedef int (*eth_representor_info_get_t)(structrte_eth_dev *dev, typedef int (*eth_rx_metadata_negotiate_t)(struct rte_eth_dev *dev,uint64_t *features); +/** +* @internal +* Get ethdev private info. Suggestion: Dump ethdev private info to a file.
Quoted 67 lines
lib/ethdev/ethdev_driver.h +11
-0+* +* @param file +* A pointer to a file for output. +* @param dev +* Port (ethdev) handle. +* +* @return +* Negative errno value on error, positive value on success. +*/ +typedef int (*eth_dev_priv_dump_t )(FILE *file,struct rte_eth_dev *dev); +/** * @internal A structure containing the functions exported by an
Ethernet driver.*/lib/ethdev/ethdev_driver.h +3
-0@@ -1186,6 +1200,9 @@ struct eth_dev_ops {* kinds of metadata to the PMD */ eth_rx_metadata_negotiate_t rx_metadata_negotiate; + +/** Dump ethdev private info */ +eth_dev_priv_dump_t eth_dev_priv_dump; };/** lib/ethdev/rte_ethdev.c +15
-0@@ -6472,6 +6472,21 @@ rte_eth_rx_metadata_negotiate(uint16_t port_id, uint64_t *features) (*dev->dev_ops->rx_metadata_negotiate)(dev, features)); } +int +rte_eth_dev_priv_dump(FILE *file, uint16_t port_id) +{ + struct rte_eth_dev *dev; + int ret; + + RTE_ETH_VALID_PORTID_OR_ERR_RET(port_id, -ENODEV); + dev = &rte_eth_devices[port_id]; + + RTE_FUNC_PTR_OR_ERR_RET(*dev->dev_ops->eth_dev_priv_dump, - ENOTSUP); + ret = (*dev->dev_ops->eth_dev_priv_dump)(file, dev); + + return ret; +} +RTE_LOG_REGISTER_DEFAULT(rte_eth_dev_logtype, INFO);
RTE_INIT(ethdev_init_telemetry)lib/ethdev/rte_ethdev.h +3
-0@@ -5888,6 +5888,22 @@ rte_eth_tx_buffer(uint16_t port_id, uint16_t queue_id, return rte_eth_tx_buffer_flush(port_id, queue_id, buffer); } +/** +* @warning +* @b EXPERIMENTAL: this API may change, or be removed, without prior Quoted 3 lines
notice
+ *
+ * Get ethdev private info.Suggestion: Dump ethdev private info to a file.
Quoted 15 lines
+ *
+ * @param file
+ * A pointer to a file for output.
+ * @paramport_id
+ * The port identifier of the Ethernet device.
+ * @return
+ * Negative errno value on error, positive value on success.
+ */lib/ethdev/rte_ethdev.h +3
-1+__rte_experimental +int rte_eth_dev_priv_dump(FILE *file, uint16_t port_id); + #ifdef __cplusplus }#endif --Quoted 2 lines
2.33.0
You should probably also add
rte_eth_dev_priv_dumpto the EXPERIMENTAL section of the version.map file.Quoted 4 lines
Quoted 4 lines
Acked-by Morten Brørup <[email protected]>.
.
RFC v3 1/1 ethdev: introduce ethdev dump API
In reply to Min Hu (Connor)Added the ethdev dump API which provides functions for query private info from device. There exists many private properties in different PMD drivers, such as adapter state, Rx/Tx func algorithm in hns3 PMD. The information of these properties is important for debug. As the information is private, the new API is introduced.
---
v3:
* fix comment.
* add rte_eth_dev_priv_dump to version.map file.
v2:
* fix dump API name
* adjust description in doc.
--- doc/guides/rel_notes/release_22_03.rst | 7 +++++++ lib/ethdev/ethdev_driver.h | 17 +++++++++++++++++ lib/ethdev/rte_ethdev.c | 15 +++++++++++++++ lib/ethdev/rte_ethdev.h | 16 ++++++++++++++++ lib/ethdev/version.map | 3 +++ 5 files changed, 58 insertions(+)
doc/guides/rel_notes/release_22_03.rst +7 -0
@@ -55,6 +55,13 @@ New Features Also, make sure to start the actual text at the margin. ======================================================= +* **Added the private ethdev dump API, for query private info of ethdev.** + + Added the private ethdev dump API which provides functions for query + private info from device. There exists many private properties in + different PMD drivers. The information of these properties is important + for debug. As the information is private, the new API is introduced. + Removed Items -------------
lib/ethdev/ethdev_driver.h +17 -0
@@ -990,6 +990,20 @@ typedef int (*eth_representor_info_get_t)(struct rte_eth_dev *dev,typedef int (*eth_rx_metadata_negotiate_t )(struct rte_eth_dev *dev,uint64_t *features); +/** +* @internal +* Dump ethdev private info to a file. +* +* @param file +* A pointer to a file for output. +* @param dev +* Port (ethdev) handle. +* +* @return +* Negative errno value on error, positive value on success. +*/ +typedef int (*eth_dev_priv_dump_t )(FILE *file,struct rte_eth_dev *dev); +/** * @internal A structure containing the functions exported by an Ethernet driver. */
@@ -1186,6 +1200,9 @@ struct eth_dev_ops {* kinds of metadata to the PMD */ eth_rx_metadata_negotiate_t rx_metadata_negotiate; + +/** Dump ethdev private info */ +eth_dev_priv_dump_t eth_dev_priv_dump; };/**
lib/ethdev/rte_ethdev.c +15 -0
@@ -6472,6 +6472,21 @@ rte_eth_rx_metadata_negotiate(uint16_t port_id, uint64_t *features) (*dev->dev_ops->rx_metadata_negotiate)(dev, features)); } +int +rte_eth_dev_priv_dump (FILE *file,uint16_t port_id) +{ +struct rte_eth_dev *dev; +int ret; + +RTE_ETH_VALID_PORTID_OR_ERR_RET (port_id, -ENODEV ); + dev = &rte_eth_devices[port_id]; + +RTE_FUNC_PTR_OR_ERR_RET (*dev->dev_ops->eth_dev_priv_dump, -ENOTSUP ); + ret = (*dev->dev_ops->eth_dev_priv_dump)(file, dev); + +return ret; +} +RTE_LOG_REGISTER_DEFAULT (rte_eth_dev_logtype,INFO );RTE_INIT (ethdev_init_telemetry)
lib/ethdev/rte_ethdev.h +16 -0
@@ -5888,6 +5888,22 @@ rte_eth_tx_buffer(uint16_t port_id, uint16_t queue_id,return rte_eth_tx_buffer_flush (port_id, queue_id, buffer); } +/** +* @warning +* @b EXPERIMENTAL: this API may change, or be removed, without prior notice +* +* Dump ethdev private info to a file. +* +* @param file +* A pointer to a file for output. +* @param port_id +* The port identifier of the Ethernet device. +* @return +* Negative errno value on error, positive value on success. +*/ +__rte_experimental +int rte_eth_dev_priv_dump(FILE *file, uint16_t port_id); + #ifdef __cplusplus }#endif
lib/ethdev/version.map +3 -0
@@ -256,6 +256,9 @@ EXPERIMENTAL { rte_flow_flex_item_create; rte_flow_flex_item_release; rte_flow_pick_transfer_proxy; + + # added in 22.03 + rte_eth_dev_priv_dump; }; INTERNAL {
Signature · 1 line
2.33.0
Quoted 16 lines · Min Hu (Connor)
Added the ethdev dump API which provides functions for query private info from device. There exists many private properties in different PMD drivers, such as adapter state, Rx/Tx func algorithm in hns3 PMD. The information of these properties is important for debug. As the information is private, the new API is introduced.
Signed-off-by Min Hu (Connor) <[email protected]>Acked-by Morten BrÞrup <[email protected]>---
v3:
* fix comment.
* addrte_eth_dev_priv_dumpto version.map file.v2:
* fix dump API name
* adjust description in doc.
Quoted 8 lines
--- doc/guides/rel_notes/release_22_03.rst | 7 +++++++ lib/ethdev/ethdev_driver.h | 17 +++++++++++++++++ lib/ethdev/rte_ethdev.c | 15 +++++++++++++++ lib/ethdev/rte_ethdev.h | 16 ++++++++++++++++ lib/ethdev/version.map | 3 +++ 5 files changed, 58 insertions(+)
Ajit Khaparde
Acked-byQuoted 29 lines · Ray Kinsella
Quoted 16 lines · Min Hu (Connor)
Added the ethdev dump API which provides functions for query private info from device. There exists many private properties in different PMD drivers, such as adapter state, Rx/Tx func algorithm in hns3 PMD. The information of these properties is important for debug. As the information is private, the new API is introduced.
Signed-off-by Min Hu (Connor) <[email protected]>Acked-by Morten BrÞrup <[email protected]>---
v3:
* fix comment.
* addrte_eth_dev_priv_dumpto version.map file.v2:
* fix dump API name
* adjust description in doc.Quoted 8 lines
--- doc/guides/rel_notes/release_22_03.rst | 7 +++++++ lib/ethdev/ethdev_driver.h | 17 +++++++++++++++++ lib/ethdev/rte_ethdev.c | 15 +++++++++++++++ lib/ethdev/rte_ethdev.h | 16 ++++++++++++++++ lib/ethdev/version.map | 3 +++ 5 files changed, 58 insertions(+)
Quoted 1 line
Acked-by Ray Kinsella <[email protected]>
Quoted 30 lines · Ajit Khaparde
Quoted 29 lines · Ray Kinsella
Quoted 24 lines · Min Hu (Connor)
Added the ethdev dump API which provides functions for query private info from device. There exists many private properties in different PMD drivers, such as adapter state, Rx/Tx func algorithm in hns3 PMD. The information of these properties is important for debug. As the information is private, the new API is introduced.
Signed-off-by Min Hu (Connor) <[email protected]>Acked-by Morten BrÞrup <[email protected]>---
v3:
* fix comment.
* addrte_eth_dev_priv_dumpto version.map file.v2:
* fix dump API name
* adjust description in doc.--- doc/guides/rel_notes/release_22_03.rst | 7 +++++++ lib/ethdev/ethdev_driver.h | 17 +++++++++++++++++ lib/ethdev/rte_ethdev.c | 15 +++++++++++++++ lib/ethdev/rte_ethdev.h | 16 ++++++++++++++++ lib/ethdev/version.map | 3 +++ 5 files changed, 58 insertions(+)
Quoted 2 lines
Quoted 1 line
Acked-by Ray Kinsella <[email protected]>Acked-by Ajit Khaparde <[email protected]>
I have a concern and this "private info", it can be useful for debug but still it will lead to a PMD specific applications, I wonder if some common information can be provided. Is there any list in your mind what can be part of this private info?
Also why not use existing xstats or telemetry to get more data from the drivers?
As synced with Ciara (cc'ed), she mentioned more information can be get via registering callbacks to ethdev.
Quoted 40 lines · Ferruh Yigit
Quoted 30 lines · Ajit Khaparde
Quoted 29 lines · Ray Kinsella
Quoted 24 lines · Min Hu (Connor)
Added the ethdev dump API which provides functions for query private info from device. There exists many private properties in different PMD drivers, such as adapter state, Rx/Tx func algorithm in hns3 PMD. The information of these properties is important for debug. As the information is private, the new API is introduced.
Signed-off-by Min Hu (Connor) <[email protected]>Acked-by Morten BrÞrup <[email protected]>---
v3:
* fix comment.
* addrte_eth_dev_priv_dumpto version.map file.v2:
* fix dump API name
* adjust description in doc.
---
doc/guides/rel_notes/release_22_03.rst| 7 +++++++ lib/ethdev/ethdev_driver.h | 17 +++++++++++++++++ lib/ethdev/rte_ethdev.c | 15 +++++++++++++++ lib/ethdev/rte_ethdev.h | 16 ++++++++++++++++ lib/ethdev/version.map | 3 +++5 files changed, 58 insertions(+)Quoted 2 lines
Quoted 1 line
Acked-by Ray Kinsella <[email protected]>Acked-by Ajit Khaparde <[email protected]>I have a concern and this "private info", it can be useful for debug but still it will lead to a PMD specific applications, I wonder if some common information can be provided. Is there any list in your mind what can be part of this private info?
Quoted 4 lines
Also why not use existing xstats or telemetry to get more data from the drivers?
As synced with Ciara (cc'ed), she mentioned more information can be get via registering callbacks to ethdev.
Opps, Ciara (Power) cc'ed now.
Hi, Ferruh,
Quoted 32 lines · Ferruh Yigit
Quoted 31 lines · Ferruh Yigit
Quoted 30 lines · Ajit Khaparde
Quoted 29 lines · Ray Kinsella
Quoted 24 lines · Min Hu (Connor)
Added the ethdev dump API which provides functions for query private info from device. There exists many private properties in different PMD drivers, such as adapter state, Rx/Tx func algorithm in hns3 PMD. The information of these properties is important for debug. As the information is private, the new API is introduced.
Signed-off-by Min Hu (Connor) <[email protected]>Acked-by Morten BrÞrup <[email protected]>---
v3:
* fix comment.
* addrte_eth_dev_priv_dumpto version.map file.v2:
* fix dump API name
* adjust description in doc.
---
doc/guides/rel_notes/release_22_03.rst| 7 +++++++ lib/ethdev/ethdev_driver.h | 17 +++++++++++++++++ lib/ethdev/rte_ethdev.c | 15 +++++++++++++++ lib/ethdev/rte_ethdev.h | 16 ++++++++++++++++ lib/ethdev/version.map | 3 +++5 files changed, 58 insertions(+)
Quoted 7 lines
Quoted 7 lines
Quoted 2 lines
Quoted 1 line
Acked-by Ray Kinsella <[email protected]>Acked-by Ajit Khaparde <[email protected]>I have a concern and this "private info", it can be useful for debug but still it will lead to a PMD specific applications, I wonder if some common information can be provided.
The purpose of introducing the API is to enhance dump ability for NICs. It will be used in APP proc-info, I will add "show-port-private" option for it. As for common information, I will add them in "show-port" option which already exists.
Is there any list in your mind what
Quoted 1 line
Quoted 1 line
can be part of this private info?
Private info for NICs, take HNS3 PMD NICS as an example,
- Dev Capability:
-- support DCB: no
-- support COPPER: no
-- support FD QUEUE REGION: no
-- support PTP: no
-- support TX PUSH: no
-- support INDEP TXRX: no
-- support STASH: no
-- support SIMPLE BD: no
-- support RXD Advanced Layout: no
-- support OUTER UDP CKSUM: no
-- support RAS IMP: no
-- support TM: no
-- support VF VLAN FILTER MOD: no
- VLAN Config Info:
-- Port VLAN filter configuration
nic_ingress :Disable
nic_egress :Disable
-- VF VLAN filter configuration
nic_ingress :Disable
nic_egress :Disable
-- RX VLAN configuration
vlan1_strip_en :Disable
vlan2_strip_en :Disable
vlan1_vlan_prionly :Disable
vlan2_vlan_prionly :Disable
vlan1_strip_discard :Disable
vlan2_strip_discard :Disable
-- TX VLAN configuration
accept_tag1 :Enable
accept_untag1 :Enable
insert_tag1_en :Disable
default_vlan_tag1 = 0, qos = 0
accept_tag2 :Enable
accept_untag2 :Enable
insert_tag2_en :Disable
default_vlan_tag2 = 0, qos = 0
vlan_shift_mode :Disable
-- pvid status: off
- Fdir Info:
-- mode=0 max_key_len=400 rule_num:512 cnt_num:32
-- key_sel=1 tuple_active=0x3bdfd890 meta_data_active=0xd0
-- ipv6_word_en: in_s=3 in_d=3 out_s=0 out_d=0....
These info is private for hns3 PMD.
Quoted 3 lines
Quoted 3 lines
Also why not use existing xstats or telemetry to get more data from the drivers?
Xstats is used to query statistics rather than state. Like above info, many keys is "enable" or "disable". It is not appropriated to put the date in xstats.
Telemetry is used to query common info for NICs, not for private info.
It is not also appropriated to put private date in telemetry.
Quoted 5 lines
Quoted 2 lines
As synced with Ciara (cc'ed), she mentioned more information can be get via registering callbacks to ethdev.
Opps, Ciara (Power) cc'ed now.
.
Quoted 39 lines · Min Hu (Connor)
Hi, Ferruh,
Quoted 36 lines · Ferruh Yigit
Quoted 35 lines · Ferruh Yigit
Quoted 32 lines · Ajit Khaparde
Quoted 29 lines · Ray Kinsella
Quoted 22 lines · Min Hu (Connor)
Added the ethdev dump API which provides functions for query private info from device. There exists many private properties in different PMD drivers, such as adapter state, Rx/Tx func algorithm in hns3 PMD. The information of these properties is important for debug. As the information is private, the new API is introduced.
Signed-off-by Min Hu (Connor) <[email protected]>Acked-by Morten BrÞrup <[email protected]>---
v3:
* fix comment.
* addrte_eth_dev_priv_dumpto version.map file.v2:
* fix dump API name
* adjust description in doc.
---
doc/guides/rel_notes/release_22_03.rst| 7 +++++++ lib/ethdev/ethdev_driver.h | 17 +++++++++++++++++ lib/ethdev/rte_ethdev.c | 15 +++++++++++++++ lib/ethdev/rte_ethdev.h | 16 ++++++++++++++++ lib/ethdev/version.map | 3 +++Quoted 2 lines
5 files changed, 58 insertions(+)Quoted 1 line
Acked-by Ray Kinsella <[email protected]>Acked-by Ajit Khaparde <[email protected]>
Quoted 8 lines
Quoted 3 lines
Quoted 3 lines
I have a concern and this "private info", it can be useful for debug but still it will lead to a PMD specific applications, I wonder if some common information can be provided.
The purpose of introducing the API is to enhance dump ability for NICs. It will be used in APP proc-info, I will add "show-port-private" option for it. As for common information, I will add them in "show-port" option which already exists.
Hi Connor,
We briefly touch on this patch in today's release status meeting, as far as I get general consensus was this can be good for debugging issues offline (instead of application parsing the output of API).
So can you please send a new, non-RFC, version of the patch? And please keep the existing ack/review tags.
Quoted 69 lines
Is there any list in your mind what
Quoted 1 line
Quoted 1 line
can be part of this private info?
Private info for NICs, take HNS3 PMD NICS as an example,
- Dev Capability:
-- support DCB: no
-- support COPPER: no
-- support FD QUEUE REGION: no
-- support PTP: no
-- support TX PUSH: no
-- support INDEP TXRX: no
-- support STASH: no
-- support SIMPLE BD: no
-- support RXD Advanced Layout: no
-- support OUTER UDP CKSUM: no
-- support RAS IMP: no
-- support TM: no
-- support VF VLAN FILTER MOD: no
- VLAN Config Info:
-- Port VLAN filter configuration
nic_ingress :Disable
nic_egress :Disable
-- VF VLAN filter configuration
nic_ingress :Disable
nic_egress :Disable
-- RX VLAN configuration
vlan1_strip_en :Disable vlan2_strip_en :Disable vlan1_vlan_prionly :Disable vlan2_vlan_prionly :Disable vlan1_strip_discard :Disable vlan2_strip_discard :Disable
-- TX VLAN configuration
accept_tag1 :Enable accept_untag1 :Enable insert_tag1_en :Disabledefault_vlan_tag1= 0, qos = 0
accept_tag2 :Enable accept_untag2 :Enable insert_tag2_en :Disabledefault_vlan_tag2= 0, qos = 0 vlan_shift_mode :Disable
-- pvid status: off
- Fdir Info:
-- mode=0 max_key_len=400 rule_num:512 cnt_num:32 -- key_sel=1 tuple_active=0x3bdfd890 meta_data_active=0xd0 --ipv6_word_en: in_s=3 in_d=3 out_s=0 out_d=0
....These info is private for hns3 PMD.
Quoted 3 lines
Quoted 3 lines
Also why not use existing xstats or telemetry to get more data from the drivers?
Xstats is used to query statistics rather than state. Like above info, many keys is "enable" or "disable". It is not appropriated to put the date in xstats.
Telemetry is used to query common info for NICs, not for private info.
It is not also appropriated to put private date in telemetry.Quoted 5 lines
Quoted 2 lines
As synced with Ciara (cc'ed), she mentioned more information can be get via registering callbacks to ethdev.
Opps, Ciara (Power) cc'ed now.
.
Quoted 49 lines · Ferruh Yigit
Quoted 39 lines · Min Hu (Connor)
Hi, Ferruh,
Quoted 36 lines · Ferruh Yigit
Quoted 35 lines · Ferruh Yigit
Quoted 32 lines · Ajit Khaparde
Quoted 29 lines · Ray Kinsella
Quoted 22 lines · Min Hu (Connor)
Added the ethdev dump API which provides functions for query private info from device. There exists many private properties in different PMD drivers, such as adapter state, Rx/Tx func algorithm in hns3 PMD. The information of these properties is important for debug. As the information is private, the new API is introduced.
Signed-off-by Min Hu (Connor) <[email protected]>Acked-by Morten BrÞrup <[email protected]>---
v3:
* fix comment.
* addrte_eth_dev_priv_dumpto version.map file.v2:
* fix dump API name
* adjust description in doc.
---
doc/guides/rel_notes/release_22_03.rst| 7 +++++++ lib/ethdev/ethdev_driver.h | 17 +++++++++++++++++ lib/ethdev/rte_ethdev.c | 15 +++++++++++++++ lib/ethdev/rte_ethdev.h | 16 ++++++++++++++++ lib/ethdev/version.map | 3 +++Quoted 2 lines
5 files changed, 58 insertions(+)Quoted 1 line
Acked-by Ray Kinsella <[email protected]>Acked-by Ajit Khaparde <[email protected]>Quoted 8 lines
Quoted 3 lines
Quoted 3 lines
I have a concern and this "private info", it can be useful for debug but still it will lead to a PMD specific applications, I wonder if some common information can be provided.
The purpose of introducing the API is to enhance dump ability for NICs. It will be used in APP proc-info, I will add "show-port-private" option for it. As for common information, I will add them in "show-port" option which already exists.
Quoted 8 lines
Hi Connor,
We briefly touch on this patch in today's release status meeting, as far as I get general consensus was this can be good for debugging issues offline (instead of application parsing the output of API).
So can you please send a new, non-RFC, version of the patch? And please keep the existing ack/review tags.
Hi, Ferruh,
Non-RFC version of the patch has been sent, please check it out.
Thanks.
Quoted 72 lines
Quoted 54 lines
Is there any list in your mind what
Quoted 1 line
Quoted 1 line
can be part of this private info?
Private info for NICs, take HNS3 PMD NICS as an example,
- Dev Capability:
-- support DCB: no
-- support COPPER: no
-- support FD QUEUE REGION: no
-- support PTP: no
-- support TX PUSH: no
-- support INDEP TXRX: no
-- support STASH: no
-- support SIMPLE BD: no
-- support RXD Advanced Layout: no
-- support OUTER UDP CKSUM: no
-- support RAS IMP: no
-- support TM: no
-- support VF VLAN FILTER MOD: no
- VLAN Config Info:
-- Port VLAN filter configuration
nic_ingress :Disable
nic_egress :Disable
-- VF VLAN filter configuration
nic_ingress :Disable
nic_egress :Disable
-- RX VLAN configuration
vlan1_strip_en :Disable vlan2_strip_en :Disable vlan1_vlan_prionly :Disable vlan2_vlan_prionly :Disable vlan1_strip_discard :Disable vlan2_strip_discard :Disable
-- TX VLAN configuration
accept_tag1 :Enable accept_untag1 :Enable insert_tag1_en :Disabledefault_vlan_tag1= 0, qos = 0
accept_tag2 :Enable accept_untag2 :Enable insert_tag2_en :Disabledefault_vlan_tag2= 0, qos = 0 vlan_shift_mode :Disable
-- pvid status: off
- Fdir Info:
-- mode=0 max_key_len=400 rule_num:512 cnt_num:32 -- key_sel=1 tuple_active=0x3bdfd890 meta_data_active=0xd0 --ipv6_word_en: in_s=3 in_d=3 out_s=0 out_d=0
....These info is private for hns3 PMD.
Quoted 1 line
Quoted 1 line
Quoted 16 lines
Quoted 2 lines
Quoted 2 lines
Also why not use existing xstats or telemetry to get more data from the drivers?
Xstats is used to query statistics rather than state. Like above info, many keys is "enable" or "disable". It is not appropriated to put the date in xstats.
Telemetry is used to query common info for NICs, not for private info. It is not also appropriated to put private date in telemetry.
Quoted 5 lines
Quoted 2 lines
As synced with Ciara (cc'ed), she mentioned more information can be get via registering callbacks to ethdev.
Opps, Ciara (Power) cc'ed now.
..
RFC v3 1/1 ethdev: introduce ethdev dump API
In reply to Min Hu (Connor)Added the ethdev dump API which provides functions for query private info from device. There exists many private properties in different PMD drivers, such as adapter state, Rx/Tx func algorithm in hns3 PMD. The information of these properties is important for debug. As the information is private, the new API is introduced.
---
v3:
* fix comment.
* add rte_eth_dev_priv_dump to version.map file.
v2:
* fix dump API name
* adjust description in doc.
--- doc/guides/rel_notes/release_22_03.rst | 7 +++++++ lib/ethdev/ethdev_driver.h | 17 +++++++++++++++++ lib/ethdev/rte_ethdev.c | 15 +++++++++++++++ lib/ethdev/rte_ethdev.h | 16 ++++++++++++++++ lib/ethdev/version.map | 3 +++ 5 files changed, 58 insertions(+)
doc/guides/rel_notes/release_22_03.rst +7 -0
@@ -55,6 +55,13 @@ New Features Also, make sure to start the actual text at the margin. ======================================================= +* **Added the private ethdev dump API, for query private info of ethdev.** + + Added the private ethdev dump API which provides functions for query + private info from device. There exists many private properties in + different PMD drivers. The information of these properties is important + for debug. As the information is private, the new API is introduced. + Removed Items -------------
lib/ethdev/ethdev_driver.h +17 -0
@@ -990,6 +990,20 @@ typedef int (*eth_representor_info_get_t)(struct rte_eth_dev *dev,typedef int (*eth_rx_metadata_negotiate_t )(struct rte_eth_dev *dev,uint64_t *features); +/** +* @internal +* Dump ethdev private info to a file. +* +* @param file +* A pointer to a file for output. +* @param dev +* Port (ethdev) handle. +* +* @return +* Negative errno value on error, positive value on success. +*/ +typedef int (*eth_dev_priv_dump_t )(FILE *file,struct rte_eth_dev *dev); +/** * @internal A structure containing the functions exported by an Ethernet driver. */
@@ -1186,6 +1200,9 @@ struct eth_dev_ops {* kinds of metadata to the PMD */ eth_rx_metadata_negotiate_t rx_metadata_negotiate; + +/** Dump ethdev private info */ +eth_dev_priv_dump_t eth_dev_priv_dump; };/**
lib/ethdev/rte_ethdev.c +15 -0
@@ -6472,6 +6472,21 @@ rte_eth_rx_metadata_negotiate(uint16_t port_id, uint64_t *features) (*dev->dev_ops->rx_metadata_negotiate)(dev, features)); } +int +rte_eth_dev_priv_dump (FILE *file,uint16_t port_id) +{ +struct rte_eth_dev *dev; +int ret; + +RTE_ETH_VALID_PORTID_OR_ERR_RET (port_id, -ENODEV ); + dev = &rte_eth_devices[port_id]; + +RTE_FUNC_PTR_OR_ERR_RET (*dev->dev_ops->eth_dev_priv_dump, -ENOTSUP ); + ret = (*dev->dev_ops->eth_dev_priv_dump)(file, dev); + +return ret; +} +RTE_LOG_REGISTER_DEFAULT (rte_eth_dev_logtype,INFO );RTE_INIT (ethdev_init_telemetry)
lib/ethdev/rte_ethdev.h +16 -0
@@ -5888,6 +5888,22 @@ rte_eth_tx_buffer(uint16_t port_id, uint16_t queue_id,return rte_eth_tx_buffer_flush (port_id, queue_id, buffer); } +/** +* @warning +* @b EXPERIMENTAL: this API may change, or be removed, without prior notice +* +* Dump ethdev private info to a file. +* +* @param file +* A pointer to a file for output. +* @param port_id +* The port identifier of the Ethernet device. +* @return +* Negative errno value on error, positive value on success. +*/ +__rte_experimental +int rte_eth_dev_priv_dump(FILE *file, uint16_t port_id); + #ifdef __cplusplus }#endif
lib/ethdev/version.map +3 -0
@@ -256,6 +256,9 @@ EXPERIMENTAL { rte_flow_flex_item_create; rte_flow_flex_item_release; rte_flow_pick_transfer_proxy; + + # added in 22.03 + rte_eth_dev_priv_dump; }; INTERNAL {
Signature · 1 line
2.33.0
PATCH 1/1 ethdev: introduce ethdev dump API
In reply to Min Hu (Connor)Added the ethdev dump API which provides functions for query private info from device. There exists many private properties in different PMD drivers, such as adapter state, Rx/Tx func algorithm in hns3 PMD. The information of these properties is important for debug. As the information is private, the new API is introduced.
--- doc/guides/rel_notes/release_22_03.rst | 7 +++++++ lib/ethdev/ethdev_driver.h | 17 +++++++++++++++++ lib/ethdev/rte_ethdev.c | 15 +++++++++++++++ lib/ethdev/rte_ethdev.h | 16 ++++++++++++++++ lib/ethdev/version.map | 3 +++ 5 files changed, 58 insertions(+)
doc/guides/rel_notes/release_22_03.rst +7 -0
@@ -92,6 +92,13 @@ New Features * Called ``rte_ipv4/6_udptcp_cksum_mbuf()`` functions in testpmd csum mode to support software UDP/TCP checksum over multiple segments. +* **Added the private ethdev dump API, for query private info of ethdev.** + + Added the private ethdev dump API which provides functions for query + private info from device. There exists many private properties in + different PMD drivers. The information of these properties is important + for debug. As the information is private, the new API is introduced. + Removed Items -------------
lib/ethdev/ethdev_driver.h +17 -0
@@ -990,6 +990,20 @@ typedef int (*eth_representor_info_get_t)(struct rte_eth_dev *dev,typedef int (*eth_rx_metadata_negotiate_t )(struct rte_eth_dev *dev,uint64_t *features); +/** +* @internal +* Dump ethdev private info to a file. +* +* @param file +* A pointer to a file for output. +* @param dev +* Port (ethdev) handle. +* +* @return +* Negative errno value on error, positive value on success. +*/ +typedef int (*eth_dev_priv_dump_t )(FILE *file,struct rte_eth_dev *dev); +/** * @internal A structure containing the functions exported by an Ethernet driver. */
@@ -1186,6 +1200,9 @@ struct eth_dev_ops {* kinds of metadata to the PMD */ eth_rx_metadata_negotiate_t rx_metadata_negotiate; + +/** Dump ethdev private info */ +eth_dev_priv_dump_t eth_dev_priv_dump; };/**
lib/ethdev/rte_ethdev.c +15 -0
@@ -6485,6 +6485,21 @@ rte_eth_rx_metadata_negotiate(uint16_t port_id, uint64_t *features) (*dev->dev_ops->rx_metadata_negotiate)(dev, features)); } +int +rte_eth_dev_priv_dump (FILE *file,uint16_t port_id) +{ +struct rte_eth_dev *dev; +int ret; + +RTE_ETH_VALID_PORTID_OR_ERR_RET (port_id, -ENODEV ); + dev = &rte_eth_devices[port_id]; + +RTE_FUNC_PTR_OR_ERR_RET (*dev->dev_ops->eth_dev_priv_dump, -ENOTSUP ); + ret = (*dev->dev_ops->eth_dev_priv_dump)(file, dev); + +return ret; +} +RTE_LOG_REGISTER_DEFAULT (rte_eth_dev_logtype,INFO );RTE_INIT (ethdev_init_telemetry)
lib/ethdev/rte_ethdev.h +16 -0
@@ -5902,6 +5902,22 @@ rte_eth_tx_buffer(uint16_t port_id, uint16_t queue_id,return rte_eth_tx_buffer_flush (port_id, queue_id, buffer); } +/** +* @warning +* @b EXPERIMENTAL: this API may change, or be removed, without prior notice +* +* Dump ethdev private info to a file. +* +* @param file +* A pointer to a file for output. +* @param port_id +* The port identifier of the Ethernet device. +* @return +* Negative errno value on error, positive value on success. +*/ +__rte_experimental +int rte_eth_dev_priv_dump(FILE *file, uint16_t port_id); + #ifdef __cplusplus }#endif
lib/ethdev/version.map +3 -0
@@ -256,6 +256,9 @@ EXPERIMENTAL { rte_flow_flex_item_create; rte_flow_flex_item_release; rte_flow_pick_transfer_proxy; + + # added in 22.03 + rte_eth_dev_priv_dump; }; INTERNAL {
Signature · 1 line
2.33.0
Quoted 1 line · Min Hu (Connor)
Added the ethdev dump API which provides functions for query private info
Isn't API and function are same thing in this contexts?
Quoted 4 lines
from device. There exists many private properties in different PMD drivers, such as adapter state, Rx/Tx func algorithm in hns3 PMD. The information of these properties is important for debug. As the information is private, the new API is introduced.>
In the patch title 'ethdev' is duplicated, can you fix it?
Quoted 18 lines
Signed-off-by Min Hu (Connor) <[email protected]>Acked-by Morten Brørup <[email protected]>Acked-by Ray Kinsella <[email protected]>Acked-by Ajit Khaparde <[email protected]>--- doc/guides/rel_notes/release_22_03.rst | 7 +++++++ lib/ethdev/ethdev_driver.h | 17 +++++++++++++++++ lib/ethdev/rte_ethdev.c | 15 +++++++++++++++ lib/ethdev/rte_ethdev.h | 16 ++++++++++++++++ lib/ethdev/version.map | 3 +++ 5 files changed, 58 insertions(+)doc/guides/rel_notes/release_22_03.rst +0
-0@@ -92,6 +92,13 @@ New Features * Called ``rte_ipv4/6_udptcp_cksum_mbuf()`` functions in testpmd csum mode
doc/guides/rel_notes/release_22_03.rst New Features · line 102
Quoted 10 lines
doc/guides/rel_notes/release_22_03.rst +7
-0to support software UDP/TCP checksum over multiple segments. +* **Added the private ethdev dump API, for query private info of ethdev.** + + Added the private ethdev dump API which provides functions for query + private info from device. There exists many private properties in + different PMD drivers. The information of these properties is important + for debug. As the information is private, the new API is introduced. +
Can you please move the update up, just above the ethdev PMD updates?
Quoted 7 lines
doc/guides/rel_notes/release_22_03.rst +0
-0Removed Items -------------lib/ethdev/ethdev_driver.h +0
-0
lib/ethdev/ethdev_driver.h typedef int (*eth_representor_info_get_t)(struct rte_eth_dev *dev, · line 996
Quoted 7 lines
lib/ethdev/ethdev_driver.h +4
-0@@ -990,6 +990,20 @@ typedef int (*eth_representor_info_get_t)(struct rte_eth_dev *dev,typedef int (*eth_rx_metadata_negotiate_t )(struct rte_eth_dev *dev,uint64_t *features); +/** +* @internal +* Dump ethdev private info to a file. +*
It doesn't dump the 'ethdev' private info, it dumps the private info from device.
Quoted 14 lines
lib/ethdev/ethdev_driver.h +10
-0+* @param file +* A pointer to a file for output. +* @param dev +* Port (ethdev) handle. +* +* @return +* Negative errno value on error, positive value on success. +*/ +typedef int (*eth_dev_priv_dump_t )(FILE *file,struct rte_eth_dev *dev); +/** * @internal A structure containing the functions exported by an Ethernet driver. */
lib/ethdev/ethdev_driver.h struct eth_dev_ops { · line 1204
Quoted 5 lines
lib/ethdev/ethdev_driver.h +2
-0@@ -1186,6 +1200,9 @@ struct eth_dev_ops {* kinds of metadata to the PMD */ eth_rx_metadata_negotiate_t rx_metadata_negotiate; + +/** Dump ethdev private info */
Ditto, not 'ethdev' private info.
Quoted 11 lines
lib/ethdev/ethdev_driver.h +1
-0+eth_dev_priv_dump_t eth_dev_priv_dump; };/** lib/ethdev/rte_ethdev.c +0
-0@@ -6485,6 +6485,21 @@ rte_eth_rx_metadata_negotiate(uint16_t port_id, uint64_t *features) (*dev->dev_ops->rx_metadata_negotiate)(dev, features)); }
lib/ethdev/rte_ethdev.c rte_eth_rx_metadata_negotiate(uint16_t port_id, uint64_t *features) · line 6496
Quoted 10 lines
lib/ethdev/rte_ethdev.c +9
-0+int +rte_eth_dev_priv_dump (FILE *file,uint16_t port_id) +{ +struct rte_eth_dev *dev; +int ret; + +RTE_ETH_VALID_PORTID_OR_ERR_RET (port_id, -ENODEV ); + dev = &rte_eth_devices[port_id]; +
What do you think to check 'file' pointer against NULL before passing it to dev_ops?
lib/ethdev/rte_ethdev.c rte_eth_rx_metadata_negotiate(uint16_t port_id, uint64_t *features) · line 6498
Quoted 2 lines
lib/ethdev/rte_ethdev.c +2
-0+RTE_FUNC_PTR_OR_ERR_RET (*dev->dev_ops->eth_dev_priv_dump, -ENOTSUP ); + ret = (*dev->dev_ops->eth_dev_priv_dump)(file, dev);
Should 'eth_err()' used here?
Quoted 12 lines
lib/ethdev/rte_ethdev.c +4
-0+ +return ret; +} +RTE_LOG_REGISTER_DEFAULT (rte_eth_dev_logtype,INFO );RTE_INIT (ethdev_init_telemetry)lib/ethdev/rte_ethdev.h +0
-0
lib/ethdev/rte_ethdev.h rte_eth_tx_buffer(uint16_t port_id, uint16_t queue_id, · line 5910
Quoted 9 lines
lib/ethdev/rte_ethdev.h +6
-0@@ -5902,6 +5902,22 @@ rte_eth_tx_buffer(uint16_t port_id, uint16_t queue_id,return rte_eth_tx_buffer_flush (port_id, queue_id, buffer); } +/** +* @warning +* @b EXPERIMENTAL: this API may change, or be removed, without prior notice +* +* Dump ethdev private info to a file. +*
Ditto.
Should we highlight that provided data and the order depends on the PMD?
lib/ethdev/rte_ethdev.h rte_eth_tx_buffer(uint16_t port_id, uint16_t queue_id, · line 5916
Quoted 6 lines
lib/ethdev/rte_ethdev.h +6
-0+* @param file +* A pointer to a file for output. +* @param port_id +* The port identifier of the Ethernet device. +* @return +* Negative errno value on error, positive value on success.
What does 'errno' mean here, does the API cause errno to be set and return it?
Also it is possible to list possible specific error values, as done with other API documentation, like:
(0) if successful.
(-ENODEV) if *port_id* is invalid.
(-ENOTSUP) if hardware doesn't support.
....
lib/ethdev/rte_ethdev.h rte_eth_tx_buffer(uint16_t port_id, uint16_t queue_id, · line 5920
Quoted 4 lines
lib/ethdev/rte_ethdev.h +4
-0+*/ +__rte_experimental +int rte_eth_dev_priv_dump(FILE *file, uint16_t port_id); +
What do you think to have the 'port_id' as first argument to be consistent with the other APIs?
Quoted 17 lines
lib/ethdev/rte_ethdev.h +0
-0#ifdef __cplusplus }#endif lib/ethdev/version.map +3
-0@@ -256,6 +256,9 @@ EXPERIMENTAL { rte_flow_flex_item_create; rte_flow_flex_item_release; rte_flow_pick_transfer_proxy; + + # added in 22.03 + rte_eth_dev_priv_dump; }; INTERNAL {
Quoted 24 lines
From: Ferruh Yigit [mailto:[email protected]]
Sent: Monday, 7 February 2022 12.46Quoted 1 line · Min Hu (Connor)
Added the ethdev dump API which provides functions for query private
info
Isn't API and function are same thing in this contexts?
Quoted 1 line
from device. There exists many private properties in different PMD
drivers,
Quoted 1 line
such as adapter state, Rx/Tx func algorithm in hns3 PMD. The
information of
Quoted 1 line
these properties is important for debug. As the information is
private,
Quoted 1 line
the new API is introduced.>
In the patch title 'ethdev' is duplicated, can you fix it?
Quoted 4 lines
Signed-off-by Min Hu (Connor) <[email protected]>Acked-by Morten Brørup <[email protected]>Acked-by Ray Kinsella <[email protected]>Acked-by Ajit Khaparde <[email protected]>
[...]
Quoted 5 lines
Review typedef int (*eth_representor_info_get_t)(struct
Quoted 1 line
(quoted diff) +0
-0@@ -990,6 +990,20 @@ typedef int (*eth_representor_info_get_t)(struct
rte_eth_dev*dev,Quoted 3 lines
lib/ethdev/ethdev_driver.h +0
-0typedef int (*eth_rx_metadata_negotiate_t )(struct rte_eth_dev *dev,uint64_t *features);
Quoted 7 lines
lib/ethdev/ethdev_driver.h typedef int (*eth_representor_info_get_t)(struct rte_eth_dev *dev, · line 996
Quoted 4 lines
lib/ethdev/ethdev_driver.h +4
-0+/** +* @internal +* Dump ethdev private info to a file. +* It doesn't dump the 'ethdev' private info, it dumps the private info from device.
It seems perfectly clear to me. How would you prefer it phrased instead?
[...]
Quoted 9 lines
lib/ethdev/rte_ethdev.h rte_eth_tx_buffer(uint16_t port_id, uint16_t queue_id, · line 5920
Quoted 5 lines
lib/ethdev/rte_ethdev.h +4
-0+*/ +__rte_experimental +int rte_eth_dev_priv_dump(FILE *file, uint16_t port_id); +What do you think to have the '
port_id' as first argument to be consistent
with the other APIs?
The _dump APIs in other libraries have the file pointer as the first parameter, so let's follow that convention here too. No need to move the port_id parameter here.
Only rte_dma_dump() has the file pointer last, and I didn't catch it when the function was defined.
Quoted 37 lines · Morten Brørup
Quoted 24 lines
From: Ferruh Yigit [mailto:[email protected]]
Sent: Monday, 7 February 2022 12.46Quoted 1 line · Min Hu (Connor)
Added the ethdev dump API which provides functions for query private
info
Isn't API and function are same thing in this contexts?
Quoted 1 line
from device. There exists many private properties in different PMD
drivers,
Quoted 1 line
such as adapter state, Rx/Tx func algorithm in hns3 PMD. The
information of
Quoted 1 line
these properties is important for debug. As the information is
private,
Quoted 1 line
the new API is introduced.>
In the patch title 'ethdev' is duplicated, can you fix it?
Quoted 4 lines
Signed-off-by Min Hu (Connor) <[email protected]>Acked-by Morten Brørup <[email protected]>Acked-by Ray Kinsella <[email protected]>Acked-by Ajit Khaparde <[email protected]>[...]
Quoted 10 lines
Review typedef int (*eth_representor_info_get_t)(struct
Quoted 1 line
(quoted diff) +0
-0@@ -990,6 +990,20 @@ typedef int (*eth_representor_info_get_t)(struct
rte_eth_dev*dev,Quoted 7 lines
lib/ethdev/ethdev_driver.h +4
-0typedef int (*eth_rx_metadata_negotiate_t )(struct rte_eth_dev *dev,uint64_t *features); +/** +* @internal +* Dump ethdev private info to a file. +*
Quoted 5 lines
Quoted 2 lines
It doesn't dump the 'ethdev' private info, it dumps the private info from device.
It seems perfectly clear to me. How would you prefer it phrased instead?
What described in the document is more accurate,
"query private info from device".
What we are dumping here is not ethdev private info, it is device private info, and we really don't know what that data may be in the ethdev layer.
Also there is a chance that 'ethdev private info' can be confused with 'ethdev->data->dev_private'
Quoted 12 lines
[...]
Quoted 9 lines
lib/ethdev/rte_ethdev.h rte_eth_tx_buffer(uint16_t port_id, uint16_t queue_id, · line 5920
Quoted 5 lines
+ */
lib/ethdev/rte_ethdev.h +3
-0+__rte_experimental +int rte_eth_dev_priv_dump(FILE *file, uint16_t port_id); +What do you think to have the '
port_id' as first argument to be consistent
with the other APIs?
Quoted 2 lines
The
_dumpAPIs in other libraries have the file pointer as the first parameter, so let's follow that convention here too. No need to move theport_idparameter here.
Yes, for most of the _dump() APIs, file pointer seems is the first argument, bu they are from various libraries.
Within the ethdev APIs, I think it makes sense that all APIs start with 'port_id' parameter for consistency, like done in:rte_flow_dev_dump(uint16_t port_id, ...)
Quoted 2 lines
Only
rte_dma_dump()has the file pointer last, and I didn't catch it when the function was defined.
Quoted 54 lines
From: Ferruh Yigit [mailto:[email protected]]
Sent: Monday, 7 February 2022 13.36Quoted 5 lines · Morten Brørup
Quoted 5 lines
From: Ferruh Yigit [mailto:[email protected]]
Sent: Monday, 7 February 2022 12.46Quoted 1 line · Min Hu (Connor)
Added the ethdev dump API which provides functions for query
private
Quoted 15 lines
Quoted 15 lines
info
Isn't API and function are same thing in this contexts?
Quoted 1 line
from device. There exists many private properties in different PMD
drivers,
Quoted 1 line
such as adapter state, Rx/Tx func algorithm in hns3 PMD. The
information of
Quoted 1 line
these properties is important for debug. As the information is
private,
Quoted 1 line
the new API is introduced.>
In the patch title 'ethdev' is duplicated, can you fix it?
Quoted 8 lines
Quoted 4 lines
Quoted 4 lines
Signed-off-by Min Hu (Connor) <[email protected]>Acked-by Morten Brørup <[email protected]>Acked-by Ray Kinsella <[email protected]>Acked-by Ajit Khaparde <[email protected]>[...]
Quoted 1 line
Review typedef int
Quoted 1 line
(quoted diff) +0
-0@@ -990,6 +990,20 @@ typedef int(*eth_representor_info_get_t)(struct
Quoted 2 lines
Quoted 2 lines
rte_eth_dev*dev,Quoted 1 line
(quoted diff) +0
-0typedef int (*eth_rx_metadata_negotiate_t)(struct rte_eth_dev*dev,
Quoted 11 lines
Quoted 9 lines
lib/ethdev/ethdev_driver.h typedef int (*eth_representor_info_get_t)(struct rte_eth_dev *dev, · line 996
Quoted 6 lines
lib/ethdev/ethdev_driver.h +4
-0uint64_t *features); +/** +* @internal +* Dump ethdev private info to a file. +* It doesn't dump the 'ethdev' private info, it dumps the private info from device.
It seems perfectly clear to me. How would you prefer it phrased
instead?
Quoted 1 line
What described in the document is more accurate,
"query private info from device".
Quoted 6 lines
What we are dumping here is not ethdev private info, it is device private info,
and we really don't know what that data may be in the ethdev layer.Also there is a chance that 'ethdev private info' can be confused with '
ethdev->data->dev_private'
OK. Now I got your point! The difference is very subtle.
Quoted 22 lines
Quoted 13 lines
[...]
Quoted 9 lines
lib/ethdev/rte_ethdev.h rte_eth_tx_buffer(uint16_t port_id, uint16_t queue_id, · line 5920
Quoted 5 lines
+ */
lib/ethdev/rte_ethdev.h +3
-0+__rte_experimental +int rte_eth_dev_priv_dump(FILE *file, uint16_t port_id); +What do you think to have the '
port_id' as first argument to be consistent
with the other APIs?Quoted 1 line
The
_dumpAPIs in other libraries have the file pointer as the firstparameter, so let's follow that convention here too. No need to move the
port_idparameter here.Quoted 1 line
Yes, for most of the
_dump()APIs, file pointer seems is the first argument,
bu they are from various libraries.
Quoted 7 lines
Within the ethdev APIs, I think it makes sense that all APIs start with '
port_id' parameter for consistency, like done in:rte_flow_dev_dump(uint16_t port_id, ...)Quoted 1 line
Only
rte_dma_dump()has the file pointer last, and I didn't catch itwhen the function was defined.
Quoted 1 line
OK. Then I agree with you about following the convention like rte_flow_dev_dump() with the port_id first.
I even think Connor got it right the first time, and I proposed following the other convention.
It's not easy when there are two opposite conventions. :-)
-Morten
Quoted 89 lines · Morten Brørup
Quoted 54 lines
From: Ferruh Yigit [mailto:[email protected]]
Sent: Monday, 7 February 2022 13.36Quoted 5 lines · Morten Brørup
Quoted 5 lines
From: Ferruh Yigit [mailto:[email protected]]
Sent: Monday, 7 February 2022 12.46Quoted 1 line · Min Hu (Connor)
Added the ethdev dump API which provides functions for query
private
Quoted 15 lines
Quoted 15 lines
info
Isn't API and function are same thing in this contexts?
Quoted 1 line
from device. There exists many private properties in different PMD
drivers,
Quoted 1 line
such as adapter state, Rx/Tx func algorithm in hns3 PMD. The
information of
Quoted 1 line
these properties is important for debug. As the information is
private,
Quoted 1 line
the new API is introduced.>
In the patch title 'ethdev' is duplicated, can you fix it?
Quoted 8 lines
Quoted 4 lines
Quoted 4 lines
Signed-off-by Min Hu (Connor) <[email protected]>Acked-by Morten Brørup <[email protected]>Acked-by Ray Kinsella <[email protected]>Acked-by Ajit Khaparde <[email protected]>[...]
Quoted 1 line
Review typedef int
Quoted 1 line
(quoted diff) +0
-0@@ -990,6 +990,20 @@ typedef int(*eth_representor_info_get_t)(struct
Quoted 2 lines
Quoted 2 lines
rte_eth_dev*dev,Quoted 1 line
(quoted diff) +0
-0typedef int (*eth_rx_metadata_negotiate_t)(struct rte_eth_dev*dev,
Quoted 11 lines
Quoted 9 lines
lib/ethdev/ethdev_driver.h typedef int (*eth_representor_info_get_t)(struct rte_eth_dev *dev, · line 996
Quoted 6 lines
lib/ethdev/ethdev_driver.h +4
-0uint64_t *features); +/** +* @internal +* Dump ethdev private info to a file. +* It doesn't dump the 'ethdev' private info, it dumps the private info from device.
It seems perfectly clear to me. How would you prefer it phrased
instead?
Quoted 1 line
What described in the document is more accurate,
"query private info from device".Quoted 6 lines
What we are dumping here is not ethdev private info, it is device private info,
and we really don't know what that data may be in the ethdev layer.Also there is a chance that 'ethdev private info' can be confused with '
ethdev->data->dev_private'OK. Now I got your point! The difference is very subtle.
Quoted 26 lines
Quoted 13 lines
[...]
Quoted 9 lines
lib/ethdev/rte_ethdev.h rte_eth_tx_buffer(uint16_t port_id, uint16_t queue_id, · line 5920
Quoted 5 lines
+ */
lib/ethdev/rte_ethdev.h +3
-0+__rte_experimental +int rte_eth_dev_priv_dump(FILE *file, uint16_t port_id); +What do you think to have the '
port_id' as first argument to be consistent
with the other APIs?Quoted 1 line
The
_dumpAPIs in other libraries have the file pointer as the firstparameter, so let's follow that convention here too. No need to move the
port_idparameter here.Quoted 1 line
Yes, for most of the
_dump()APIs, file pointer seems is the first argument,
bu they are from various libraries.Within the ethdev APIs, I think it makes sense that all APIs start with '
port_id' parameter for consistency, like done in:rte_flow_dev_dump(uint16_t port_id, ...)
Quoted 8 lines
Quoted 3 lines
Quoted 1 line
Only
rte_dma_dump()has the file pointer last, and I didn't catch itwhen the function was defined.
Quoted 1 line
OK. Then I agree with you about following the convention like
rte_flow_dev_dump()with theport_idfirst.I even think Connor got it right the first time, and I proposed following the other convention.
Ahh, may bad I missed that, sorry for not commenting on time.
Quoted 2 lines
It's not easy when there are two opposite conventions. :-)
Yep, that is the main issue.
Hi, Ferruh,
Quoted 49 lines · Ferruh Yigit
Quoted 48 lines · Morten Brørup
Quoted 48 lines
From: Ferruh Yigit [mailto:[email protected]]
Sent: Monday, 7 February 2022 13.36Quoted 5 lines · Morten Brørup
Quoted 5 lines
From: Ferruh Yigit [mailto:[email protected]]
Sent: Monday, 7 February 2022 12.46Quoted 1 line · Min Hu (Connor)
Added the ethdev dump API which provides functions for query
private
Quoted 15 lines
Quoted 15 lines
info
Isn't API and function are same thing in this contexts?
Quoted 1 line
from device. There exists many private properties in different PMD
drivers,
Quoted 1 line
such as adapter state, Rx/Tx func algorithm in hns3 PMD. The
information of
Quoted 1 line
these properties is important for debug. As the information is
private,
Quoted 1 line
the new API is introduced.>
In the patch title 'ethdev' is duplicated, can you fix it?
Quoted 8 lines
Quoted 4 lines
Quoted 4 lines
Signed-off-by Min Hu (Connor) <[email protected]>Acked-by Morten Brørup <[email protected]>Acked-by Ray Kinsella <[email protected]>Acked-by Ajit Khaparde <[email protected]>[...]
Quoted 1 line
Review typedef int
Quoted 1 line
(quoted diff) +0
-0@@ -990,6 +990,20 @@ typedef int(*eth_representor_info_get_t)(struct
Quoted 2 lines
Quoted 2 lines
rte_eth_dev*dev,Quoted 1 line
typedef int (*eth_rx_metadata_negotiate_t)(struct
rte_eth_dev*dev,
Quoted 11 lines
Quoted 9 lines
Quoted 6 lines
uint64_t*features);+/**
+ * @internal
+ * Dump ethdev private info to a file.
+ *It doesn't dump the 'ethdev' private info, it dumps the private info from device.
It seems perfectly clear to me. How would you prefer it phrased
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
instead?
Quoted 1 line
What described in the document is more accurate,
"query private info from device".What we are dumping here is not ethdev private info, it is device private info,
what is the difference between ethdev and device?
Quoted 4 lines
Quoted 4 lines
Quoted 4 lines
and we really don't know what that data may be in the ethdev layer.
Also there is a chance that 'ethdev private info' can be confused with '
ethdev->data->dev_private'
what I want to dump is exactly the 'ethdev->data->dev_private'. 'ethdev private info' means 'ethdev->data->dev_private'. why confused?
Quoted 47 lines
Quoted 29 lines
OK. Now I got your point! The difference is very subtle.
Quoted 26 lines
Quoted 13 lines
[...]
Quoted 9 lines
lib/ethdev/rte_ethdev.h rte_eth_tx_buffer(uint16_t port_id, uint16_t queue_id, · line 5920
Quoted 5 lines
+ */
lib/ethdev/rte_ethdev.h +3
-0+__rte_experimental +int rte_eth_dev_priv_dump(FILE *file, uint16_t port_id); +What do you think to have the '
port_id' as first argument to be consistent
with the other APIs?Quoted 1 line
The
_dumpAPIs in other libraries have the file pointer as the firstparameter, so let's follow that convention here too. No need to move the
port_idparameter here.Quoted 1 line
Yes, for most of the
_dump()APIs, file pointer seems is the first argument,
bu they are from various libraries.Within the ethdev APIs, I think it makes sense that all APIs start with '
port_id' parameter for consistency, like done in:rte_flow_dev_dump(uint16_t port_id, ...)Quoted 8 lines
Quoted 3 lines
Quoted 1 line
Only
rte_dma_dump()has the file pointer last, and I didn't catch itwhen the function was defined.
Quoted 1 line
OK. Then I agree with you about following the convention like
rte_flow_dev_dump()with theport_idfirst.I even think Connor got it right the first time, and I proposed following the other convention.
Ahh, may bad I missed that, sorry for not commenting on time.
Quoted 2 lines
It's not easy when there are two opposite conventions. :-)
Yep, that is the main issue.
.
Quoted 55 lines · Min Hu (Connor)
Hi, Ferruh,
Quoted 52 lines · Ferruh Yigit
Quoted 51 lines · Morten Brørup
Quoted 51 lines
From: Ferruh Yigit [mailto:[email protected]]
Sent: Monday, 7 February 2022 13.36Quoted 5 lines · Morten Brørup
Quoted 5 lines
From: Ferruh Yigit [mailto:[email protected]]
Sent: Monday, 7 February 2022 12.46Quoted 1 line · Min Hu (Connor)
Added the ethdev dump API which provides functions for query
private
Quoted 15 lines
Quoted 15 lines
info
Isn't API and function are same thing in this contexts?
Quoted 1 line
from device. There exists many private properties in different PMD
drivers,
Quoted 1 line
such as adapter state, Rx/Tx func algorithm in hns3 PMD. The
information of
Quoted 1 line
these properties is important for debug. As the information is
private,
Quoted 1 line
the new API is introduced.>
In the patch title 'ethdev' is duplicated, can you fix it?
Quoted 8 lines
Quoted 4 lines
Quoted 4 lines
Signed-off-by Min Hu (Connor) <[email protected]>Acked-by Morten Brørup <[email protected]>Acked-by Ray Kinsella <[email protected]>Acked-by Ajit Khaparde <[email protected]>[...]
Quoted 1 line
Review typedef int
Quoted 1 line
(quoted diff) +0
-0@@ -990,6 +990,20 @@ typedef int(*eth_representor_info_get_t)(struct
Quoted 2 lines
Quoted 2 lines
rte_eth_dev*dev,Quoted 1 line
typedef int (*eth_rx_metadata_negotiate_t)(struct
rte_eth_dev*dev,
Quoted 11 lines
Quoted 9 lines
Quoted 6 lines
uint64_t*features);+/**
+ * @internal
+ * Dump ethdev private info to a file.
+ *It doesn't dump the 'ethdev' private info, it dumps the private info from device.
It seems perfectly clear to me. How would you prefer it phrased
instead?
Quoted 1 line
Quoted 7 lines
Quoted 5 lines
Quoted 5 lines
Quoted 5 lines
What described in the document is more accurate,
"query private info from device".What we are dumping here is not ethdev private info, it is device private info,
what is the difference between ethdev and device?
It is not very clear, but for me 'ethdev' is refers to device abstract layer (ethdev library) specific private data and device refers to ethdev device (PMD) private data. ethdev is common for all drivers.
Quoted 7 lines
Quoted 4 lines
Quoted 4 lines
Quoted 4 lines
and we really don't know what that data may be in the ethdev layer.
Also there is a chance that 'ethdev private info' can be confused with '
ethdev->data->dev_private'what I want to dump is exactly the '
ethdev->data->dev_private'. 'ethdev private info' means 'ethdev->data->dev_private'. why confused?
What I understand was this API can return any device private information, it is not limited to 'ethdev->data->dev_private', (although most of the data is represented in this struct), like if you want to dump queue state, this is out of 'ethdev->data->dev_private'.
Quoted 47 lines
Quoted 47 lines
Quoted 29 lines
OK. Now I got your point! The difference is very subtle.
Quoted 26 lines
Quoted 13 lines
[...]
Quoted 9 lines
Review
Quoted 5 lines
+ */
(quoted diff) +3
-0+__rte_experimental +int rte_eth_dev_priv_dump(FILE *file, uint16_t port_id); +What do you think to have the '
port_id' as first argument to be consistent
with the other APIs?Quoted 1 line
The
_dumpAPIs in other libraries have the file pointer as the firstparameter, so let's follow that convention here too. No need to move the
port_idparameter here.Quoted 1 line
Yes, for most of the
_dump()APIs, file pointer seems is the first argument,
bu they are from various libraries.Within the ethdev APIs, I think it makes sense that all APIs start with '
port_id' parameter for consistency, like done in:rte_flow_dev_dump(uint16_t port_id, ...)Quoted 8 lines
Quoted 3 lines
Quoted 1 line
Only
rte_dma_dump()has the file pointer last, and I didn't catch itwhen the function was defined.
Quoted 1 line
OK. Then I agree with you about following the convention like
rte_flow_dev_dump()with theport_idfirst.I even think Connor got it right the first time, and I proposed following the other convention.
Ahh, may bad I missed that, sorry for not commenting on time.
Quoted 2 lines
It's not easy when there are two opposite conventions. :-)
Yep, that is the main issue.
.
Hi, Ferruh,
Quoted 59 lines · Ferruh Yigit
Quoted 58 lines · Min Hu (Connor)
Hi, Ferruh,
Quoted 55 lines · Ferruh Yigit
Quoted 54 lines · Morten Brørup
Quoted 54 lines
From: Ferruh Yigit [mailto:[email protected]]
Sent: Monday, 7 February 2022 13.36Quoted 5 lines · Morten Brørup
Quoted 5 lines
From: Ferruh Yigit [mailto:[email protected]]
Sent: Monday, 7 February 2022 12.46Quoted 1 line · Min Hu (Connor)
Added the ethdev dump API which provides functions for query
private
Quoted 15 lines
Quoted 15 lines
info
Isn't API and function are same thing in this contexts?
Quoted 1 line
from device. There exists many private properties in different PMD
drivers,
Quoted 1 line
such as adapter state, Rx/Tx func algorithm in hns3 PMD. The
information of
Quoted 1 line
these properties is important for debug. As the information is
private,
Quoted 1 line
the new API is introduced.>
In the patch title 'ethdev' is duplicated, can you fix it?
Quoted 8 lines
Quoted 4 lines
Quoted 4 lines
Signed-off-by Min Hu (Connor) <[email protected]>Acked-by Morten Brørup <[email protected]>Acked-by Ray Kinsella <[email protected]>Acked-by Ajit Khaparde <[email protected]>[...]
Quoted 1 line
Review typedef int
Quoted 1 line
(quoted diff) +0
-0@@ -990,6 +990,20 @@ typedef int(*eth_representor_info_get_t)(struct
Quoted 2 lines
Quoted 2 lines
rte_eth_dev*dev,Quoted 1 line
typedef int (*eth_rx_metadata_negotiate_t)(struct
rte_eth_dev*dev,
Quoted 11 lines
Quoted 9 lines
Quoted 6 lines
uint64_t*features);+/**
+ * @internal
+ * Dump ethdev private info to a file.
+ *It doesn't dump the 'ethdev' private info, it dumps the private info from device.
It seems perfectly clear to me. How would you prefer it phrased
instead?
Quoted 1 line
What described in the document is more accurate,
"query private info from device".
Quoted 7 lines
Quoted 4 lines
Quoted 2 lines
Quoted 2 lines
Quoted 2 lines
What we are dumping here is not ethdev private info, it is device private info,
what is the difference between ethdev and device?
It is not very clear, but for me 'ethdev' is refers to device abstract layer (ethdev library) specific private data
Could you give an example for 'ethdev'specific private data ?
and device refers to ethdev
Quoted 1 line
device (PMD) private data. ethdev is common for all drivers.
OK, we could treat it as convention in future.
Quoted 11 lines
Quoted 8 lines
Quoted 5 lines
Quoted 5 lines
Quoted 5 lines
and we really don't know what that data may be in the ethdev layer.
Also there is a chance that 'ethdev private info' can be confused with '
ethdev->data->dev_private'what I want to dump is exactly the '
ethdev->data->dev_private'. 'ethdev private info' means 'ethdev->data->dev_private'. why confused?What I understand was this API can return any device private information, it is not limited to '
ethdev->data->dev_private', (although most of the
I think this API is limited to 'ethdev->data->dev_private'.
Quoted 3 lines
data
is represented in this struct), like if you want to dump queue state, this is out of 'ethdev->data->dev_private'.
Queue state can be dumped using API 'rte_eth_tx_queue_info_get'.
Quoted 50 lines
Quoted 42 lines
Quoted 42 lines
Quoted 30 lines
OK. Now I got your point! The difference is very subtle.
Quoted 26 lines
Quoted 13 lines
[...]
Quoted 9 lines
Review
Quoted 5 lines
+ */
(quoted diff) +3
-0+__rte_experimental +int rte_eth_dev_priv_dump(FILE *file, uint16_t port_id); +What do you think to have the '
port_id' as first argument to be consistent
with the other APIs?Quoted 1 line
The
_dumpAPIs in other libraries have the file pointer as the firstparameter, so let's follow that convention here too. No need to move the
port_idparameter here.Quoted 1 line
Yes, for most of the
_dump()APIs, file pointer seems is the first argument,
bu they are from various libraries.Within the ethdev APIs, I think it makes sense that all APIs start with '
port_id' parameter for consistency, like done in:rte_flow_dev_dump(uint16_t port_id, ...)Quoted 8 lines
Quoted 3 lines
Quoted 1 line
Only
rte_dma_dump()has the file pointer last, and I didn't catch itwhen the function was defined.
Quoted 1 line
OK. Then I agree with you about following the convention like
rte_flow_dev_dump()with theport_idfirst.I even think Connor got it right the first time, and I proposed following the other convention.
Ahh, may bad I missed that, sorry for not commenting on time.
Quoted 6 lines
Quoted 6 lines
Quoted 2 lines
It's not easy when there are two opposite conventions. :-)
Yep, that is the main issue.
.
.
Quoted 62 lines · Min Hu (Connor)
Hi, Ferruh,
Quoted 59 lines · Ferruh Yigit
Quoted 58 lines · Min Hu (Connor)
Hi, Ferruh,
Quoted 55 lines · Ferruh Yigit
Quoted 54 lines · Morten Brørup
Quoted 54 lines
From: Ferruh Yigit [mailto:[email protected]]
Sent: Monday, 7 February 2022 13.36Quoted 5 lines · Morten Brørup
Quoted 5 lines
From: Ferruh Yigit [mailto:[email protected]]
Sent: Monday, 7 February 2022 12.46Quoted 1 line · Min Hu (Connor)
Added the ethdev dump API which provides functions for query
private
Quoted 15 lines
Quoted 15 lines
info
Isn't API and function are same thing in this contexts?
Quoted 1 line
from device. There exists many private properties in different PMD
drivers,
Quoted 1 line
such as adapter state, Rx/Tx func algorithm in hns3 PMD. The
information of
Quoted 1 line
these properties is important for debug. As the information is
private,
Quoted 1 line
the new API is introduced.>
In the patch title 'ethdev' is duplicated, can you fix it?
Quoted 8 lines
Quoted 4 lines
Quoted 4 lines
Signed-off-by Min Hu (Connor) <[email protected]>Acked-by Morten Brørup <[email protected]>Acked-by Ray Kinsella <[email protected]>Acked-by Ajit Khaparde <[email protected]>[...]
Quoted 1 line
Review typedef int
Quoted 1 line
(quoted diff) +0
-0@@ -990,6 +990,20 @@ typedef int(*eth_representor_info_get_t)(struct
Quoted 2 lines
Quoted 2 lines
rte_eth_dev*dev,Quoted 1 line
typedef int (*eth_rx_metadata_negotiate_t)(struct
rte_eth_dev*dev,
Quoted 11 lines
Quoted 9 lines
Quoted 6 lines
uint64_t*features);+/**
+ * @internal
+ * Dump ethdev private info to a file.
+ *It doesn't dump the 'ethdev' private info, it dumps the private info from device.
It seems perfectly clear to me. How would you prefer it phrased
instead?
Quoted 1 line
What described in the document is more accurate,
"query private info from device".
Quoted 9 lines
Quoted 7 lines
Quoted 4 lines
Quoted 2 lines
Quoted 2 lines
Quoted 2 lines
What we are dumping here is not ethdev private info, it is device private info,
what is the difference between ethdev and device?
It is not very clear, but for me 'ethdev' is refers to device abstract layer (ethdev library) specific private data
Could you give an example for 'ethdev'specific private data ?
I think 'struct rte_eth_dev' content can be a sample.
But I hear you, diff is not clear, it is subtle as Morten said, when doc and commit log refers it as "private info from device", I think we can use the same in the API documentation as well.
Quoted 12 lines
and device refers to ethdev
Quoted 1 line
device (PMD) private data. ethdev is common for all drivers.
OK, we could treat it as convention in future.
Quoted 9 lines
Quoted 8 lines
Quoted 5 lines
Quoted 5 lines
Quoted 5 lines
and we really don't know what that data may be in the ethdev layer.
Also there is a chance that 'ethdev private info' can be confused with '
ethdev->data->dev_private'what I want to dump is exactly the '
ethdev->data->dev_private'. 'ethdev private info' means 'ethdev->data->dev_private'. why confused?
Quoted 8 lines
Quoted 2 lines
What I understand was this API can return any device private information, it is not limited to '
ethdev->data->dev_private', (although most of theI think this API is limited to '
ethdev->data->dev_private'.Quoted 3 lines
data
is represented in this struct), like if you want to dump queue state, this is out of 'ethdev->data->dev_private'.Queue state can be dumped using API '
rte_eth_tx_queue_info_get'.
Yes it can be. But as far as I can see there is nothing prevents the dump() API to provide the same, it is up to PMD.
If the intention is to limit what can be dump to 'ethdev->data->dev_private', it is not clear from API documentation/implementation.
Quoted 50 lines
Quoted 50 lines
Quoted 42 lines
Quoted 42 lines
Quoted 30 lines
OK. Now I got your point! The difference is very subtle.
Quoted 26 lines
Quoted 13 lines
[...]
Quoted 9 lines
Review
Quoted 5 lines
+ */
(quoted diff) +3
-0+__rte_experimental +int rte_eth_dev_priv_dump(FILE *file, uint16_t port_id); +What do you think to have the '
port_id' as first argument to be consistent
with the other APIs?Quoted 1 line
The
_dumpAPIs in other libraries have the file pointer as the firstparameter, so let's follow that convention here too. No need to move the
port_idparameter here.Quoted 1 line
Yes, for most of the
_dump()APIs, file pointer seems is the first argument,
bu they are from various libraries.Within the ethdev APIs, I think it makes sense that all APIs start with '
port_id' parameter for consistency, like done in:rte_flow_dev_dump(uint16_t port_id, ...)Quoted 8 lines
Quoted 3 lines
Quoted 1 line
Only
rte_dma_dump()has the file pointer last, and I didn't catch itwhen the function was defined.
Quoted 1 line
OK. Then I agree with you about following the convention like
rte_flow_dev_dump()with theport_idfirst.I even think Connor got it right the first time, and I proposed following the other convention.
Ahh, may bad I missed that, sorry for not commenting on time.
Quoted 6 lines
Quoted 6 lines
Quoted 2 lines
It's not easy when there are two opposite conventions. :-)
Yep, that is the main issue.
.
.
08/02/2022 13:59, Ferruh Yigit:
Quoted 8 lines · Min Hu (Connor)
Quoted 7 lines · Ferruh Yigit
Quoted 2 lines
What I understand was this API can return any device private information, it is not limited to '
ethdev->data->dev_private', (although most of theI think this API is limited to '
ethdev->data->dev_private'.Quoted 3 lines
data
is represented in this struct), like if you want to dump queue state, this is out of 'ethdev->data->dev_private'.
Quoted 8 lines
Quoted 2 lines
Queue state can be dumped using API '
rte_eth_tx_queue_info_get'.Yes it can be. But as far as I can see there is nothing prevents the
dump()API to provide the same, it is up to PMD.If the intention is to limit what can be dump to '
ethdev->data->dev_private', it is not clear from API documentation/implementation.
Why limiting?
The dump could even print debug infos which are nowhere else.
Quoted 12 lines · Thomas Monjalon
08/02/2022 13:59, Ferruh Yigit:
Quoted 11 lines · Min Hu (Connor)
Quoted 9 lines · Ferruh Yigit
Quoted 2 lines
What I understand was this API can return any device private information, it is not limited to '
ethdev->data->dev_private', (although most of theI think this API is limited to '
ethdev->data->dev_private'.Quoted 3 lines
data
is represented in this struct), like if you want to dump queue state, this is out of 'ethdev->data->dev_private'.Queue state can be dumped using API '
rte_eth_tx_queue_info_get'.
Quoted 8 lines
Quoted 5 lines
Yes it can be. But as far as I can see there is nothing prevents the
dump()API to provide the same, it is up to PMD.If the intention is to limit what can be dump to '
ethdev->data->dev_private', it is not clear from API documentation/implementation.Why limiting?
The dump could even print debug infos which are nowhere else.
Yes, no need to limit, it is up to PMD, just as Ferruh said.
Quoted 5 lines
.
Hi,
Quoted 70 lines · Ferruh Yigit
Quoted 69 lines · Min Hu (Connor)
Hi, Ferruh,
Quoted 66 lines · Ferruh Yigit
Quoted 55 lines · Min Hu (Connor)
Hi, Ferruh,
Quoted 52 lines · Ferruh Yigit
Quoted 51 lines · Morten Brørup
Quoted 51 lines
From: Ferruh Yigit [mailto:[email protected]]
Sent: Monday, 7 February 2022 13.36Quoted 5 lines · Morten Brørup
Quoted 5 lines
From: Ferruh Yigit [mailto:[email protected]]
Sent: Monday, 7 February 2022 12.46Quoted 1 line · Min Hu (Connor)
Added the ethdev dump API which provides functions for query
private
Quoted 15 lines
Quoted 15 lines
info
Isn't API and function are same thing in this contexts?
Quoted 1 line
from device. There exists many private properties in different PMD
drivers,
Quoted 1 line
such as adapter state, Rx/Tx func algorithm in hns3 PMD. The
information of
Quoted 1 line
these properties is important for debug. As the information is
private,
Quoted 1 line
the new API is introduced.>
In the patch title 'ethdev' is duplicated, can you fix it?
Quoted 8 lines
Quoted 4 lines
Quoted 4 lines
Signed-off-by Min Hu (Connor) <[email protected]>Acked-by Morten Brørup <[email protected]>Acked-by Ray Kinsella <[email protected]>Acked-by Ajit Khaparde <[email protected]>[...]
Quoted 1 line
Review typedef int
Quoted 1 line
(quoted diff) +0
-0@@ -990,6 +990,20 @@ typedef int(*eth_representor_info_get_t)(struct
Quoted 2 lines
Quoted 2 lines
rte_eth_dev*dev,Quoted 1 line
typedef int (*eth_rx_metadata_negotiate_t)(struct
rte_eth_dev*dev,
Quoted 11 lines
Quoted 9 lines
Quoted 6 lines
uint64_t*features);+/**
+ * @internal
+ * Dump ethdev private info to a file.
+ *It doesn't dump the 'ethdev' private info, it dumps the private info from device.
It seems perfectly clear to me. How would you prefer it phrased
instead?
Quoted 1 line
Quoted 7 lines
Quoted 5 lines
Quoted 5 lines
Quoted 5 lines
What described in the document is more accurate,
"query private info from device".What we are dumping here is not ethdev private info, it is device private info,
what is the difference between ethdev and device?
It is not very clear, but for me 'ethdev' is refers to device abstract layer (ethdev library) specific private data
Quoted 9 lines
Quoted 2 lines
Could you give an example for 'ethdev'specific private data ?
I think 'struct
rte_eth_dev' content can be a sample.But I hear you, diff is not clear, it is subtle as Morten said, when doc and commit log refers it as "private info from device", I think we can use the same in the API documentation as well.
Agreed, I will fix it in release_22_03.rst.
Quoted 79 lines
Quoted 12 lines
and device refers to ethdev
Quoted 1 line
device (PMD) private data. ethdev is common for all drivers.
OK, we could treat it as convention in future.
Quoted 9 lines
Quoted 8 lines
Quoted 5 lines
Quoted 5 lines
Quoted 5 lines
and we really don't know what that data may be in the ethdev layer.
Also there is a chance that 'ethdev private info' can be confused with '
ethdev->data->dev_private'what I want to dump is exactly the '
ethdev->data->dev_private'. 'ethdev private info' means 'ethdev->data->dev_private'. why confused?Quoted 8 lines
Quoted 2 lines
What I understand was this API can return any device private information, it is not limited to '
ethdev->data->dev_private', (although most of theI think this API is limited to '
ethdev->data->dev_private'.Quoted 3 lines
data
is represented in this struct), like if you want to dump queue state, this is out of 'ethdev->data->dev_private'.Queue state can be dumped using API '
rte_eth_tx_queue_info_get'.Yes it can be. But as far as I can see there is nothing prevents the
dump()API to provide the same, it is up to PMD.If the intention is to limit what can be dump to '
ethdev->data->dev_private', it is not clear from API documentation/implementation.Quoted 42 lines
Quoted 42 lines
Quoted 42 lines
Quoted 42 lines
Quoted 30 lines
OK. Now I got your point! The difference is very subtle.
Quoted 26 lines
Quoted 13 lines
[...]
Quoted 9 lines
Review
Quoted 5 lines
+ */
(quoted diff) +3
-0+__rte_experimental +int rte_eth_dev_priv_dump(FILE *file, uint16_t port_id); +What do you think to have the '
port_id' as first argument to be consistent
with the other APIs?Quoted 1 line
The
_dumpAPIs in other libraries have the file pointer as the firstparameter, so let's follow that convention here too. No need to move the
port_idparameter here.Quoted 1 line
Yes, for most of the
_dump()APIs, file pointer seems is the first argument,
bu they are from various libraries.Within the ethdev APIs, I think it makes sense that all APIs start with '
port_id' parameter for consistency, like done in:rte_flow_dev_dump(uint16_t port_id, ...)Quoted 8 lines
Quoted 3 lines
Quoted 1 line
Only
rte_dma_dump()has the file pointer last, and I didn't catch itwhen the function was defined.
Quoted 1 line
OK. Then I agree with you about following the convention like
rte_flow_dev_dump()with theport_idfirst.I even think Connor got it right the first time, and I proposed following the other convention.
Ahh, may bad I missed that, sorry for not commenting on time.
Quoted 8 lines
Quoted 8 lines
Quoted 6 lines
Quoted 6 lines
Quoted 2 lines
It's not easy when there are two opposite conventions. :-)
Yep, that is the main issue.
.
.
.
HI, Ferruh,
v2 has been sent according comments, please check it, thanks.
Quoted 182 lines · Ferruh Yigit
Quoted 1 line · Min Hu (Connor)
Added the ethdev dump API which provides functions for query private info
Isn't API and function are same thing in this contexts?
Quoted 4 lines
from device. There exists many private properties in different PMD drivers, such as adapter state, Rx/Tx func algorithm in hns3 PMD. The information of these properties is important for debug. As the information is private, the new API is introduced.>
In the patch title 'ethdev' is duplicated, can you fix it?
Quoted 20 lines
Signed-off-by Min Hu (Connor) <[email protected]>Acked-by Morten Brørup <[email protected]>Acked-by Ray Kinsella <[email protected]>Acked-by Ajit Khaparde <[email protected]>---
doc/guides/rel_notes/release_22_03.rst| 7 +++++++ lib/ethdev/ethdev_driver.h | 17 +++++++++++++++++ lib/ethdev/rte_ethdev.c | 15 +++++++++++++++ lib/ethdev/rte_ethdev.h | 16 ++++++++++++++++ lib/ethdev/version.map | 3 +++5 files changed, 58 insertions(+)doc/guides/rel_notes/release_22_03.rst +0
-0@@ -92,6 +92,13 @@ New Features * Called ``rte_ipv4/6_udptcp_cksum_mbuf()`` functions in testpmd csum modeto support software UDP/TCP checksum over multiple segments. +* **Added the private ethdev dump API, for query private info of
Quoted 8 lines
ethdev.**
+
+ Added the private ethdev dump API which provides functions for query + private info from device. There exists many private properties in + different PMD drivers. The information of these properties is important
+ for debug. As the information is private, the new API is introduced. +Can you please move the update up, just above the ethdev PMD updates?
Quoted 7 lines
Removed Items
-------------lib/ethdev/ethdev_driver.h +0
-0lib/ethdev/ethdev_driver.h typedef int (*eth_representor_info_get_t)(struct · line 996
Quoted 7 lines
lib/ethdev/ethdev_driver.h +4
-0@@ -990,6 +990,20 @@ typedef int (*eth_representor_info_get_t)(structrte_eth_dev *dev, typedef int (*eth_rx_metadata_negotiate_t)(struct rte_eth_dev *dev, uint64_t *features); +/** +* @internal +* Dump ethdev private info to a file. +* It doesn't dump the 'ethdev' private info, it dumps the private info from device.
Quoted 14 lines
lib/ethdev/ethdev_driver.h +10
-0+* @param file +* A pointer to a file for output. +* @param dev +* Port (ethdev) handle. +* +* @return +* Negative errno value on error, positive value on success. +*/ +typedef int (*eth_dev_priv_dump_t )(FILE *file,struct rte_eth_dev *dev); + /*** @internal A structure containing the functions exported by an Ethernet driver. */ lib/ethdev/ethdev_driver.h struct eth_dev_ops { · line 1204
Quoted 5 lines
lib/ethdev/ethdev_driver.h +2
-0@@ -1186,6 +1200,9 @@ struct eth_dev_ops {* kinds of metadata to the PMD */ eth_rx_metadata_negotiate_t rx_metadata_negotiate; + + /** Dump ethdev private info */Ditto, not 'ethdev' private info.
Quoted 10 lines
lib/ethdev/ethdev_driver.h +1
-0+ eth_dev_priv_dump_t eth_dev_priv_dump; }; /**lib/ethdev/rte_ethdev.c +0
-0@@ -6485,6 +6485,21 @@ rte_eth_rx_metadata_negotiate(uint16_t port_id, uint64_t *features) (*dev->dev_ops->rx_metadata_negotiate)(dev, features));lib/ethdev/rte_ethdev.c rte_eth_rx_metadata_negotiate(uint16_t port_id, · line 6496
Quoted 10 lines
lib/ethdev/rte_ethdev.c +9
-0} +int +rte_eth_dev_priv_dump (FILE *file,uint16_t port_id) +{ + struct rte_eth_dev *dev; + int ret; + + RTE_ETH_VALID_PORTID_OR_ERR_RET(port_id, -ENODEV); + dev = &rte_eth_devices[port_id]; +What do you think to check 'file' pointer against NULL before passing it to
dev_ops?lib/ethdev/rte_ethdev.c rte_eth_rx_metadata_negotiate(uint16_t port_id, uint64_t *features) · line 6498
Quoted 2 lines
lib/ethdev/rte_ethdev.c +2
-0+ RTE_FUNC_PTR_OR_ERR_RET(*dev->dev_ops->eth_dev_priv_dump, -ENOTSUP); + ret = (*dev->dev_ops->eth_dev_priv_dump)(file, dev);Should '
eth_err()' used here?Quoted 17 lines
lib/ethdev/rte_ethdev.c +4
-0+ + return ret; +} + RTE_LOG_REGISTER_DEFAULT(rte_eth_dev_logtype, INFO); RTE_INIT(ethdev_init_telemetry)lib/ethdev/rte_ethdev.h +3
-0@@ -5902,6 +5902,22 @@ rte_eth_tx_buffer(uint16_t port_id, uint16_t queue_id, return rte_eth_tx_buffer_flush(port_id, queue_id, buffer); } +/** +* @warning +* @b EXPERIMENTAL: this API may change, or be removed, without prior Quoted 4 lines
notice
+ *
+ * Dump ethdev private info to a file.
+ *Ditto.
Should we highlight that provided data and the order depends on the PMD?
Quoted 6 lines
+ * @param file
+ * A pointer to a file for output.
+ * @paramport_id
+ * The port identifier of the Ethernet device.
+ * @return
+ * Negative errno value on error, positive value on success.What does 'errno' mean here, does the API cause errno to be set and return it?
Also it is possible to list possible specific error values, as done with other API documentation, like:
(0) if successful.
(-ENODEV) if *port_id* is invalid.
(-ENOTSUP) if hardware doesn't support.
....lib/ethdev/rte_ethdev.h rte_eth_tx_buffer(uint16_t port_id, uint16_t queue_id, · line 5920
Quoted 4 lines
+ */
lib/ethdev/rte_ethdev.h +3
-0+__rte_experimental +int rte_eth_dev_priv_dump(FILE *file, uint16_t port_id); +What do you think to have the '
port_id' as first argument to be consistent with the other APIs?Quoted 8 lines
#ifdef
__cplusplus
}
#endiflib/ethdev/version.map +0
-0lib/ethdev/version.map EXPERIMENTAL { · line 263
Quoted 8 lines
lib/ethdev/version.map +3
-0@@ -256,6 +256,9 @@ EXPERIMENTAL { rte_flow_flex_item_create; rte_flow_flex_item_release; rte_flow_pick_transfer_proxy; + + # added in 22.03 + rte_eth_dev_priv_dump; }; INTERNAL {.
PATCH v2 1/1 ethdev: introduce dump API
In reply to Min Hu (Connor)Added the ethdev dump API which provides querying private info from ethdev. There exists many private properties in different PMD drivers, such as adapter state, Rx/Tx func algorithm in hns3 PMD. The information of these properties is important for debug. As the information is private, the new API is introduced.
---
v2:
* fixed comments from Ferruh.
--- doc/guides/rel_notes/release_22_03.rst | 7 +++++++ lib/ethdev/ethdev_driver.h | 23 +++++++++++++++++++++++ lib/ethdev/rte_ethdev.c | 17 +++++++++++++++++ lib/ethdev/rte_ethdev.h | 20 ++++++++++++++++++++ lib/ethdev/version.map | 3 +++ 5 files changed, 70 insertions(+)
doc/guides/rel_notes/release_22_03.rst +7 -0
@@ -64,6 +64,13 @@ New Features - ``rte_ipv6_udptcp_cksum_mbuf()`` - ``rte_ipv6_udptcp_cksum_mbuf_verify()`` +* **Added the private ethdev dump API, for query private info of ethdev.** + + Added the private ethdev dump API which provides functions for query + private info from device. There exists many private properties in + different PMD drivers. The information of these properties is important + for debug. As the information is private, the new API is introduced. + * **Updated AF_XDP PMD** * Added support for libxdp >=v1.2.2.
lib/ethdev/ethdev_driver.h +23 -0
@@ -990,6 +990,26 @@ typedef int (*eth_representor_info_get_t)(struct rte_eth_dev *dev,typedef int (*eth_rx_metadata_negotiate_t )(struct rte_eth_dev *dev,uint64_t *features); +/** +* @internal +* Dump private info from ethdev to a file. +* +* @param dev +* Port (ethdev) handle. +* @param file +* A pointer to a file for output. +* +* @retval 0 +* Success +* @retval -EINVAL +* Invalid file +* @retval -ENOTSUP +* Not supported +* @retval -ENODEV +* Invalid port_id +*/ +typedef int (*eth_dev_priv_dump_t )(struct rte_eth_dev *dev,FILE *file); +/** * @internal A structure containing the functions exported by an Ethernet driver. */
@@ -1186,6 +1206,9 @@ struct eth_dev_ops {* kinds of metadata to the PMD */ eth_rx_metadata_negotiate_t rx_metadata_negotiate; + +/** Dump private info from ethdev */ +eth_dev_priv_dump_t eth_dev_priv_dump; };/**
lib/ethdev/rte_ethdev.c +17 -0
@@ -6485,6 +6485,23 @@ rte_eth_rx_metadata_negotiate(uint16_t port_id, uint64_t *features) (*dev->dev_ops->rx_metadata_negotiate)(dev, features)); } +int +rte_eth_dev_priv_dump (uint16_t port_id,FILE *file) +{ +struct rte_eth_dev *dev; + +RTE_ETH_VALID_PORTID_OR_ERR_RET (port_id, -ENODEV ); + dev = &rte_eth_devices[port_id]; + +if (file ==NULL ) { +RTE_ETHDEV_LOG (ERR ,"Invalid file (NULL)\n" ); +return -EINVAL ; + } + +RTE_FUNC_PTR_OR_ERR_RET (*dev->dev_ops->eth_dev_priv_dump, -ENOTSUP ); +return eth_err (port_id, (*dev->dev_ops->eth_dev_priv_dump)(dev, file)); +} +RTE_LOG_REGISTER_DEFAULT (rte_eth_dev_logtype,INFO );RTE_INIT (ethdev_init_telemetry)
lib/ethdev/rte_ethdev.h +20 -0
@@ -5902,6 +5902,26 @@ rte_eth_tx_buffer(uint16_t port_id, uint16_t queue_id,return rte_eth_tx_buffer_flush (port_id, queue_id, buffer); } +/** +* @warning +* @b EXPERIMENTAL: this API may change, or be removed, without prior notice +* +* Dump private info from ethdev to a file. Provided data and the order depends +* on the PMD. +* +* @param port_id +* The port identifier of the Ethernet device. +* @param file +* A pointer to a file for output. +* @return +* - (-ENODEV) if *port_id* is invalid. +* - (-EINVAL) if null file. +* - (-ENOTSUP) if the device does not support this function. +* - (0) on success +*/ +__rte_experimental +int rte_eth_dev_priv_dump(uint16_t port_id, FILE *file); + #ifdef __cplusplus }#endif
lib/ethdev/version.map +3 -0
@@ -256,6 +256,9 @@ EXPERIMENTAL { rte_flow_flex_item_create; rte_flow_flex_item_release; rte_flow_pick_transfer_proxy; + + # added in 22.03 + rte_eth_dev_priv_dump; }; INTERNAL {
Signature · 1 line
2.33.0
PATCH v3 1/1 ethdev: introduce dump API
In reply to Min Hu (Connor)Added the ethdev dump API which provides querying private info from ethdev. There exists many private properties in different PMD drivers, such as adapter state, Rx/Tx func algorithm in hns3 PMD. The information of these properties is important for debug. As the information is private, the new API is introduced.
---
v3:
* change 'ethdev' to 'device'
v2:
* fixed comments from Ferruh.
--- doc/guides/rel_notes/release_22_03.rst | 7 +++++++ lib/ethdev/ethdev_driver.h | 23 +++++++++++++++++++++++ lib/ethdev/rte_ethdev.c | 17 +++++++++++++++++ lib/ethdev/rte_ethdev.h | 20 ++++++++++++++++++++ lib/ethdev/version.map | 3 +++ 5 files changed, 70 insertions(+)
doc/guides/rel_notes/release_22_03.rst +7 -0
@@ -64,6 +64,13 @@ New Features - ``rte_ipv6_udptcp_cksum_mbuf()`` - ``rte_ipv6_udptcp_cksum_mbuf_verify()`` +* **Added the private dump API, for query private info from device.** + + Added the private dump API which provides querying private info from device. + There exists many private properties in different PMD drivers. The + information of these properties is important for debug. As the information + is private, the new API is introduced. + * **Updated AF_XDP PMD** * Added support for libxdp >=v1.2.2.
lib/ethdev/ethdev_driver.h +23 -0
@@ -990,6 +990,26 @@ typedef int (*eth_representor_info_get_t)(struct rte_eth_dev *dev,typedef int (*eth_rx_metadata_negotiate_t )(struct rte_eth_dev *dev,uint64_t *features); +/** +* @internal +* Dump private info from device to a file. +* +* @param dev +* Port (ethdev) handle. +* @param file +* A pointer to a file for output. +* +* @retval 0 +* Success +* @retval -EINVAL +* Invalid file +* @retval -ENOTSUP +* Not supported +* @retval -ENODEV +* Invalid port_id +*/ +typedef int (*eth_dev_priv_dump_t )(struct rte_eth_dev *dev,FILE *file); +/** * @internal A structure containing the functions exported by an Ethernet driver. */
@@ -1186,6 +1206,9 @@ struct eth_dev_ops {* kinds of metadata to the PMD */ eth_rx_metadata_negotiate_t rx_metadata_negotiate; + +/** Dump private info from device */ +eth_dev_priv_dump_t eth_dev_priv_dump; };/**
lib/ethdev/rte_ethdev.c +17 -0
@@ -6485,6 +6485,23 @@ rte_eth_rx_metadata_negotiate(uint16_t port_id, uint64_t *features) (*dev->dev_ops->rx_metadata_negotiate)(dev, features)); } +int +rte_eth_dev_priv_dump (uint16_t port_id,FILE *file) +{ +struct rte_eth_dev *dev; + +RTE_ETH_VALID_PORTID_OR_ERR_RET (port_id, -ENODEV ); + dev = &rte_eth_devices[port_id]; + +if (file ==NULL ) { +RTE_ETHDEV_LOG (ERR ,"Invalid file (NULL)\n" ); +return -EINVAL ; + } + +RTE_FUNC_PTR_OR_ERR_RET (*dev->dev_ops->eth_dev_priv_dump, -ENOTSUP ); +return eth_err (port_id, (*dev->dev_ops->eth_dev_priv_dump)(dev, file)); +} +RTE_LOG_REGISTER_DEFAULT (rte_eth_dev_logtype,INFO );RTE_INIT (ethdev_init_telemetry)
lib/ethdev/rte_ethdev.h +20 -0
@@ -5902,6 +5902,26 @@ rte_eth_tx_buffer(uint16_t port_id, uint16_t queue_id,return rte_eth_tx_buffer_flush (port_id, queue_id, buffer); } +/** +* @warning +* @b EXPERIMENTAL: this API may change, or be removed, without prior notice +* +* Dump private info from device to a file. Provided data and the order depends +* on the PMD. +* +* @param port_id +* The port identifier of the Ethernet device. +* @param file +* A pointer to a file for output. +* @return +* - (-ENODEV) if *port_id* is invalid. +* - (-EINVAL) if null file. +* - (-ENOTSUP) if the device does not support this function. +* - (0) on success +*/ +__rte_experimental +int rte_eth_dev_priv_dump(uint16_t port_id, FILE *file); + #ifdef __cplusplus }#endif
lib/ethdev/version.map +3 -0
@@ -256,6 +256,9 @@ EXPERIMENTAL { rte_flow_flex_item_create; rte_flow_flex_item_release; rte_flow_pick_transfer_proxy; + + # added in 22.03 + rte_eth_dev_priv_dump; }; INTERNAL {
Signature · 1 line
2.33.0
Re: [PATCH v3] ethdev: introduce dump API
Quoted 56 lines · Min Hu (Connor)
Added the ethdev dump API which provides querying private info from ethdev. There exists many private properties in different PMD drivers, such as adapter state, Rx/Tx func algorithm in hns3 PMD. The information of these properties is important for debug. As the information is private, the new API is introduced.
Signed-off-by Min Hu (Connor) <[email protected]>Acked-by Morten Brørup <[email protected]>Acked-by Ray Kinsella <[email protected]>Acked-by Ajit Khaparde <[email protected]>---
v3:
* change 'ethdev' to 'device'
v2:
* fixed comments from Ferruh.--- doc/guides/rel_notes/release_22_03.rst | 7 +++++++ lib/ethdev/ethdev_driver.h | 23 +++++++++++++++++++++++ lib/ethdev/rte_ethdev.c | 17 +++++++++++++++++ lib/ethdev/rte_ethdev.h | 20 ++++++++++++++++++++ lib/ethdev/version.map | 3 +++ 5 files changed, 70 insertions(+)doc/guides/rel_notes/release_22_03.rst +7
-0@@ -64,6 +64,13 @@ New Features - ``rte_ipv6_udptcp_cksum_mbuf()`` - ``rte_ipv6_udptcp_cksum_mbuf_verify()`` +* **Added the private dump API, for query private info from device.** + + Added the private dump API which provides querying private info from device. + There exists many private properties in different PMD drivers. The + information of these properties is important for debug. As the information + is private, the new API is introduced. + * **Updated AF_XDP PMD** * Added support for libxdp >=v1.2.2.lib/ethdev/ethdev_driver.h +7
-0@@ -990,6 +990,26 @@ typedef int (*eth_representor_info_get_t)(struct rte_eth_dev *dev,typedef int (*eth_rx_metadata_negotiate_t )(struct rte_eth_dev *dev,uint64_t *features); +/** +* @internal +* Dump private info from device to a file. +* +* @param dev +* Port (ethdev) handle. +* @param file
lib/ethdev/ethdev_driver.h typedef int (*eth_representor_info_get_t)(struct rte_eth_dev *dev, · line 1009
Quoted 10 lines
lib/ethdev/ethdev_driver.h +10
-0+* A pointer to a file for output. +* +* @retval 0 +* Success +* @retval -EINVAL +* Invalid file +* @retval -ENOTSUP +* Not supported +* @retval -ENODEV +* Invalid port_id
For dev_ops (eth_dev_priv_dump_t), I don't think '-ENOTSUP' and '-ENODEV' are valid errors, those cases covered by API before dev_ops called.
Also '@return' is missing.
What about below:
*
* @return
* Negative value on error, 0 on success.
*
* @retval 0
* Success
* @retval -EINVAL
* Invalid file
*/
Quoted 31 lines
lib/ethdev/ethdev_driver.h +6
-0+*/ +typedef int (*eth_dev_priv_dump_t )(struct rte_eth_dev *dev,FILE *file); +/** * @internal A structure containing the functions exported by an Ethernet driver. */ @@ -1186,6 +1206,9 @@ struct eth_dev_ops {* kinds of metadata to the PMD */ eth_rx_metadata_negotiate_t rx_metadata_negotiate; + +/** Dump private info from device */ +eth_dev_priv_dump_t eth_dev_priv_dump; };/** lib/ethdev/rte_ethdev.c +7
-0@@ -6485,6 +6485,23 @@ rte_eth_rx_metadata_negotiate(uint16_t port_id, uint64_t *features) (*dev->dev_ops->rx_metadata_negotiate)(dev, features)); } +int +rte_eth_dev_priv_dump (uint16_t port_id,FILE *file) +{ +struct rte_eth_dev *dev; + +RTE_ETH_VALID_PORTID_OR_ERR_RET (port_id, -ENODEV ); + dev = &rte_eth_devices[port_id];
lib/ethdev/rte_ethdev.c rte_eth_rx_metadata_negotiate(uint16_t port_id, uint64_t *features) · line 6504
Quoted 10 lines
lib/ethdev/rte_ethdev.c +10
-0+ +if (file ==NULL ) { +RTE_ETHDEV_LOG (ERR ,"Invalid file (NULL)\n" ); +return -EINVAL ; + } + +RTE_FUNC_PTR_OR_ERR_RET (*dev->dev_ops->eth_dev_priv_dump, -ENOTSUP ); +return eth_err (port_id, (*dev->dev_ops->eth_dev_priv_dump)(dev, file)); +} +
Can you please move the function up in the file?
Bottom of the file is for static inline functions, APIs are above the "#include <rte_ethdev_core.h>" line, can you move the function above that?
Quoted 17 lines
lib/ethdev/rte_ethdev.c +0
-0RTE_LOG_REGISTER_DEFAULT (rte_eth_dev_logtype,INFO );RTE_INIT (ethdev_init_telemetry)lib/ethdev/rte_ethdev.h +6
-0@@ -5902,6 +5902,26 @@ rte_eth_tx_buffer(uint16_t port_id, uint16_t queue_id,return rte_eth_tx_buffer_flush (port_id, queue_id, buffer); } +/** +* @warning +* @b EXPERIMENTAL: this API may change, or be removed, without prior notice +* +* Dump private info from device to a file. Provided data and the order depends +* on the PMD.
lib/ethdev/rte_ethdev.h rte_eth_tx_buffer(uint16_t port_id, uint16_t queue_id, · line 5920
Quoted 10 lines
lib/ethdev/rte_ethdev.h +10
-0+* +* @param port_id +* The port identifier of the Ethernet device. +* @param file +* A pointer to a file for output. +* @return +* - (-ENODEV) if *port_id* is invalid. +* - (-EINVAL) if null file. +* - (-ENOTSUP) if the device does not support this function. +* - (0) on success
Can you please list "(0) on success" to be consistent with (most :( of the) other comments?
Also need to add "(-EIO) if device is removed." too, which can be returned by 'eth_err()'.
Quoted 12 lines
lib/ethdev/rte_ethdev.h +4
-0+*/ +__rte_experimental +int rte_eth_dev_priv_dump(uint16_t port_id, FILE *file); + #ifdef __cplusplus }#endif lib/ethdev/version.map +0
-0
lib/ethdev/version.map EXPERIMENTAL { · line 261
Quoted 6 lines
lib/ethdev/version.map +3
-0@@ -256,6 +256,9 @@ EXPERIMENTAL { rte_flow_flex_item_create; rte_flow_flex_item_release; rte_flow_pick_transfer_proxy; + + # added in 22.03 + rte_eth_dev_priv_dump;
next version needs to be rebased on top of next-net, because of other APIs merged, fyi
Quoted 3 lines
lib/ethdev/version.map +0
-0}; INTERNAL {
Re: [PATCH v3] ethdev: introduce dump API
Quoted 145 lines · Ferruh Yigit
Quoted 53 lines · Min Hu (Connor)
Added the ethdev dump API which provides querying private info from ethdev. There exists many private properties in different PMD drivers, such as adapter state, Rx/Tx func algorithm in hns3 PMD. The information of these properties is important for debug. As the information is private, the new API is introduced.
Signed-off-by Min Hu (Connor) <[email protected]>Acked-by Morten Brørup <[email protected]>Acked-by Ray Kinsella <[email protected]>Acked-by Ajit Khaparde <[email protected]>---
v3:
* change 'ethdev' to 'device'
v2:
* fixed comments from Ferruh.
---
doc/guides/rel_notes/release_22_03.rst| 7 +++++++ lib/ethdev/ethdev_driver.h | 23 +++++++++++++++++++++++ lib/ethdev/rte_ethdev.c | 17 +++++++++++++++++ lib/ethdev/rte_ethdev.h | 20 ++++++++++++++++++++ lib/ethdev/version.map | 3 +++5 files changed, 70 insertions(+)doc/guides/rel_notes/release_22_03.rst +0
-0@@ -64,6 +64,13 @@ New Features- ``rte_ipv6_udptcp_cksum_mbuf()``
- ``rte_ipv6_udptcp_cksum_mbuf_verify()``doc/guides/rel_notes/release_22_03.rst +7
-0+* **Added the private dump API, for query private info from device.** + + Added the private dump API which provides querying private info from device. + There exists many private properties in different PMD drivers. The + information of these properties is important for debug. As the information + is private, the new API is introduced. +* **Updated
AF_XDPPMD**
* Added support for libxdp >=v1.2.2.lib/ethdev/ethdev_driver.h +7
-0@@ -990,6 +990,26 @@ typedef int (*eth_representor_info_get_t)(struct rte_eth_dev *dev, typedef int (*eth_rx_metadata_negotiate_t)(struct rte_eth_dev *dev, uint64_t *features); +/** +* @internal +* Dump private info from device to a file. +* +* @param dev +* Port (ethdev) handle. +* @param file lib/ethdev/ethdev_driver.h typedef int (*eth_representor_info_get_t)(struct rte_eth_dev *dev, · line 1008
Quoted 10 lines
lib/ethdev/ethdev_driver.h +10
-0+ * A pointer to a file for output. + * + * @retval 0 + * Success + * @retval -EINVAL + * Invalid file + * @retval -ENOTSUP + * Not supported + * @retval -ENODEV + * Invalid port_idFor
dev_ops(eth_dev_priv_dump_t), I don't think '-ENOTSUP' and '-ENODEV' are valid errors, those cases covered by API beforedev_opscalled.Also '@return' is missing.
What about below:
*
* @return
* Negative value on error, 0 on success.
*
* @retval 0
* Success
* @retval-EINVAL
* Invalid file
*/Quoted 29 lines
lib/ethdev/ethdev_driver.h +6
-0+*/ +typedef int (*eth_dev_priv_dump_t )(struct rte_eth_dev *dev,FILE *file); + /*** @internal A structure containing the functions exported by an Ethernet driver. */ @@ -1186,6 +1206,9 @@ struct eth_dev_ops {* kinds of metadata to the PMD */ eth_rx_metadata_negotiate_t rx_metadata_negotiate; + + /** Dump private info from device */ + eth_dev_priv_dump_t eth_dev_priv_dump; }; /**lib/ethdev/rte_ethdev.c +7
-0@@ -6485,6 +6485,23 @@ rte_eth_rx_metadata_negotiate(uint16_t port_id, uint64_t *features) (*dev->dev_ops->rx_metadata_negotiate)(dev, features)); } +int +rte_eth_dev_priv_dump (uint16_t port_id,FILE *file) +{ + struct rte_eth_dev *dev; + + RTE_ETH_VALID_PORTID_OR_ERR_RET(port_id, -ENODEV); + dev = &rte_eth_devices[port_id];lib/ethdev/rte_ethdev.c rte_eth_rx_metadata_negotiate(uint16_t port_id, uint64_t *features) · line 6503
Quoted 10 lines
lib/ethdev/rte_ethdev.c +10
-0+ + if (file == NULL) { + RTE_ETHDEV_LOG(ERR, "Invalid file (NULL)\n"); + return -EINVAL; + } + + RTE_FUNC_PTR_OR_ERR_RET(*dev->dev_ops->eth_dev_priv_dump, -ENOTSUP); + return eth_err(port_id, (*dev->dev_ops->eth_dev_priv_dump)(dev, file)); +} +Can you please move the function up in the file?
Bottom of the file is for static inline functions, APIs are above the "#include <rte_ethdev_core.h>" line, can you move the function above that?Quoted 19 lines
lib/ethdev/rte_ethdev.c +0
-0RTE_LOG_REGISTER_DEFAULT(rte_eth_dev_logtype, INFO); RTE_INIT(ethdev_init_telemetry)lib/ethdev/rte_ethdev.h +10
-0@@ -5902,6 +5902,26 @@ rte_eth_tx_buffer(uint16_t port_id, uint16_t queue_id, return rte_eth_tx_buffer_flush(port_id, queue_id, buffer); } +/** +* @warning +* @b EXPERIMENTAL: this API may change, or be removed, without prior notice +* +* Dump private info from device to a file. Provided data and the order depends +* on the PMD. +* +* @param port_id +* The port identifier of the Ethernet device. +* @param file
Quoted 9 lines
lib/ethdev/rte_ethdev.h rte_eth_tx_buffer(uint16_t port_id, uint16_t queue_id, · line 5920
Quoted 6 lines
lib/ethdev/rte_ethdev.h +6
-0+ * A pointer to a file for output. + * @return + * - (-ENODEV) if *port_id* is invalid. + * - (-EINVAL) if null file. + * - (-ENOTSUP) if the device does not support this function. + * - (0) on successCan you please list "(0) on success" to be consistent with (most :( of the) other comments?
Can you please list "(0) on success" as first item ...
* @return
* - (0) on success * - (-ENODEV) if *port_id* is invalid. * - (-EINVAL) if null file.
...
Quoted 29 lines
Also need to add "(
-EIO) if device is removed." too, which can be returned by 'eth_err()'.Quoted 12 lines
+ */
lib/ethdev/rte_ethdev.h +3
-0+__rte_experimental +int rte_eth_dev_priv_dump(uint16_t port_id, FILE *file); +#ifdef
__cplusplus
}
#endiflib/ethdev/version.map +0
-0lib/ethdev/version.map EXPERIMENTAL { · line 261
Quoted 6 lines
lib/ethdev/version.map +3
-0@@ -256,6 +256,9 @@ EXPERIMENTAL { rte_flow_flex_item_create; rte_flow_flex_item_release; rte_flow_pick_transfer_proxy; + + # added in 22.03 + rte_eth_dev_priv_dump;next version needs to be rebased on top of next-net, because of other APIs merged, fyi
Quoted 2 lines
lib/ethdev/version.map +0
-0}; INTERNAL {
Re: [PATCH v3] ethdev: introduce dump API
Hi, Ferruh,
all fixed in v4, please check it, thanks.
Quoted 205 lines · Ferruh Yigit
Quoted 155 lines · Ferruh Yigit
Quoted 56 lines · Min Hu (Connor)
Added the ethdev dump API which provides querying private info from ethdev. There exists many private properties in different PMD drivers, such as adapter state, Rx/Tx func algorithm in hns3 PMD. The information of these properties is important for debug. As the information is private, the new API is introduced.
Signed-off-by Min Hu (Connor) <[email protected]>Acked-by Morten Brørup <[email protected]>Acked-by Ray Kinsella <[email protected]>Acked-by Ajit Khaparde <[email protected]>---
v3:
* change 'ethdev' to 'device'
v2:
* fixed comments from Ferruh.
---
doc/guides/rel_notes/release_22_03.rst| 7 +++++++ lib/ethdev/ethdev_driver.h | 23 +++++++++++++++++++++++ lib/ethdev/rte_ethdev.c | 17 +++++++++++++++++ lib/ethdev/rte_ethdev.h | 20 ++++++++++++++++++++ lib/ethdev/version.map | 3 +++5 files changed, 70 insertions(+)doc/guides/rel_notes/release_22_03.rst +0
-0@@ -64,6 +64,13 @@ New Features- ``rte_ipv6_udptcp_cksum_mbuf()``
- ``rte_ipv6_udptcp_cksum_mbuf_verify()``doc/guides/rel_notes/release_22_03.rst +3
-0+* **Added the private dump API, for query private info from device.** + + Added the private dump API which provides querying private infofrom device.
+ There exists many private properties in different PMD drivers. The + information of these properties is important for debug. As the information
+ is private, the new API is introduced.
+
* **UpdatedAF_XDPPMD**
* Added support for libxdp >=v1.2.2.lib/ethdev/ethdev_driver.h +7
-0@@ -990,6 +990,26 @@ typedef int (*eth_representor_info_get_t)(structrte_eth_dev *dev, typedef int (*eth_rx_metadata_negotiate_t)(struct rte_eth_dev *dev, uint64_t *features); +/** +* @internal +* Dump private info from device to a file. +* +* @param dev +* Port (ethdev) handle. +* @param file lib/ethdev/ethdev_driver.h typedef int (*eth_representor_info_get_t)(struct · line 1009
Quoted 10 lines
lib/ethdev/ethdev_driver.h +10
-0+ * A pointer to a file for output. + * + * @retval 0 + * Success + * @retval -EINVAL + * Invalid file + * @retval -ENOTSUP + * Not supported + * @retval -ENODEV + * Invalid port_idFor
dev_ops(eth_dev_priv_dump_t), I don't think '-ENOTSUP' and '-ENODEV' are valid errors, those cases covered by API beforedev_opscalled.Also '@return' is missing.
What about below:
*
* @return
* Negative value on error, 0 on success.
*
* @retval 0
* Success
* @retval-EINVAL
* Invalid file
*/Quoted 33 lines
lib/ethdev/ethdev_driver.h +3
-0+*/ +typedef int (*eth_dev_priv_dump_t )(struct rte_eth_dev *dev,FILE *file); + /*** @internal A structure containing the functions exported by an Ethernet driver. */
lib/ethdev/ethdev_driver.h +3
-0@@ -1186,6 +1206,9 @@ struct eth_dev_ops {* kinds of metadata to the PMD */ eth_rx_metadata_negotiate_t rx_metadata_negotiate; + + /** Dump private info from device */ + eth_dev_priv_dump_t eth_dev_priv_dump; }; /**lib/ethdev/rte_ethdev.c +9
-0@@ -6485,6 +6485,23 @@ rte_eth_rx_metadata_negotiate(uint16_t port_id, uint64_t *features) (*dev->dev_ops->rx_metadata_negotiate)(dev, features)); } +int +rte_eth_dev_priv_dump (uint16_t port_id,FILE *file) +{ + struct rte_eth_dev *dev; + + RTE_ETH_VALID_PORTID_OR_ERR_RET(port_id, -ENODEV); + dev = &rte_eth_devices[port_id]; + + if (file == NULL) {lib/ethdev/rte_ethdev.c rte_eth_rx_metadata_negotiate(uint16_t · line 6505
Quoted 10 lines
lib/ethdev/rte_ethdev.c +8
-1+ RTE_ETHDEV_LOG(ERR, "Invalid file (NULL)\n"); + return -EINVAL; + } + + RTE_FUNC_PTR_OR_ERR_RET(*dev->dev_ops->eth_dev_priv_dump, -ENOTSUP); + return eth_err(port_id, (*dev->dev_ops->eth_dev_priv_dump)(dev, file)); +} +Can you please move the function up in the file?
Bottom of the file is for static inline functions, APIs are above the "#include <rte_ethdev_core.h>" line, can you move the function above that?Quoted 22 lines
lib/ethdev/rte_ethdev.c +0
-0RTE_LOG_REGISTER_DEFAULT(rte_eth_dev_logtype, INFO);
RTE_INIT(ethdev_init_telemetry)lib/ethdev/rte_ethdev.h +3
-0@@ -5902,6 +5902,26 @@ rte_eth_tx_buffer(uint16_t port_id, uint16_t queue_id, return rte_eth_tx_buffer_flush(port_id, queue_id, buffer); } +/** +* @warning +* @b EXPERIMENTAL: this API may change, or be removed, without prior notice
+ *
+ * Dump private info from device to a file. Provided data and the order depends
+ * on the PMD.
+ *
+ * @paramport_id
+ * The port identifier of the Ethernet device.
+ * @param fileQuoted 9 lines
Quoted 6 lines
+ * A pointer to a file for output.
+ * @return
+ * - (-ENODEV) if *port_id* is invalid.
+ * - (-EINVAL) if null file.
+ * - (-ENOTSUP) if the device does not support this function.
+ * - (0) on successCan you please list "(0) on success" to be consistent with (most :( of the) other comments?
Can you please list "(0) on success" as first item ...
* @return
* - (0) on success
* - (-ENODEV) if *port_id* is invalid.
* - (-EINVAL) if null file.
...Quoted 17 lines
Also need to add "(
-EIO) if device is removed." too, which can be returned by 'eth_err()'.Quoted 13 lines
+ */
lib/ethdev/rte_ethdev.h +3
-0+__rte_experimental +int rte_eth_dev_priv_dump(uint16_t port_id, FILE *file); +#ifdef
__cplusplus
}
#endiflib/ethdev/version.map +0
-0@@ -256,6 +256,9 @@ EXPERIMENTAL { rte_flow_flex_item_create;Quoted 12 lines
lib/ethdev/version.map EXPERIMENTAL { · line 261
Quoted 5 lines
lib/ethdev/version.map +3
-0rte_flow_flex_item_release; rte_flow_pick_transfer_proxy; + + # added in 22.03 + rte_eth_dev_priv_dump;next version needs to be rebased on top of next-net, because of other APIs merged, fyi
Quoted 2 lines
lib/ethdev/version.map +0
-0}; INTERNAL {.
Re: [PATCH v3] ethdev: introduce dump API
In reply to Min Hu (Connor)Quoted 15 lines · Min Hu (Connor)
Added the ethdev dump API which provides querying private info from ethdev. There exists many private properties in different PMD drivers, such as adapter state, Rx/Tx func algorithm in hns3 PMD. The information of these properties is important for debug. As the information is private, the new API is introduced.
Signed-off-by Min Hu (Connor)<[email protected]>Acked-by Morten Brørup<[email protected]>Acked-by Ray Kinsella<[email protected]>Acked-by Ajit Khaparde<[email protected]>---
v3:
* change 'ethdev' to 'device'
v2:
* fixed comments from Ferruh.
Quoted 6 lines
--- doc/guides/rel_notes/release_22_03.rst | 7 +++++++ lib/ethdev/ethdev_driver.h | 23 +++++++++++++++++++++++ lib/ethdev/rte_ethdev.c | 17 +++++++++++++++++ lib/ethdev/rte_ethdev.h | 20 ++++++++++++++++++++ lib/ethdev/version.map | 3 +++
Btw, can you please confirm that there will be a PMD implementation in this release, (it can be after -rc1)?
Because our policy requires at least one PMD implementation is implemented to accept a new API.
(Thomas/David, please correct me if I remember the policy wrong)
Re: [PATCH v3] ethdev: introduce dump API
Hi, Ferruh,
Quoted 24 lines · Ferruh Yigit
Quoted 10 lines · Min Hu (Connor)
Added the ethdev dump API which provides querying private info from ethdev. There exists many private properties in different PMD drivers, such as adapter state, Rx/Tx func algorithm in hns3 PMD. The information of these properties is important for debug. As the information is private, the new API is introduced.
Signed-off-by Min Hu (Connor)<[email protected]>Acked-by Morten Brørup<[email protected]>Acked-by Ray Kinsella<[email protected]>Acked-by Ajit Khaparde<[email protected]>Quoted 11 lines
---
v3:
* change 'ethdev' to 'device'
v2:
* fixed comments from Ferruh.
---
doc/guides/rel_notes/release_22_03.rst| 7 +++++++ lib/ethdev/ethdev_driver.h | 23 +++++++++++++++++++++++ lib/ethdev/rte_ethdev.c | 17 +++++++++++++++++ lib/ethdev/rte_ethdev.h | 20 ++++++++++++++++++++ lib/ethdev/version.map | 3 +++
Quoted 2 lines
Btw, can you please confirm that there will be a PMD implementation
in this release, (it can be after -rc1)?YES, I will send a set of patches about hns3 PMD implementation once the
API is accepted.
Quoted 4 lines
Because our policy requires at least one PMD implementation is implemented to accept a new API.
(Thomas/David, please correct me if I remember the policy wrong) .
Re: [PATCH v3] ethdev: introduce dump API
Quoted 27 lines · Min Hu (Connor)
Hi, Ferruh,
Quoted 24 lines · Ferruh Yigit
Quoted 21 lines · Min Hu (Connor)
Added the ethdev dump API which provides querying private info from ethdev. There exists many private properties in different PMD drivers, such as adapter state, Rx/Tx func algorithm in hns3 PMD. The information of these properties is important for debug. As the information is private, the new API is introduced.
Signed-off-by Min Hu (Connor)<[email protected]>Acked-by Morten Brørup<[email protected]>Acked-by Ray Kinsella<[email protected]>Acked-by Ajit Khaparde<[email protected]>---
v3:
* change 'ethdev' to 'device'
v2:
* fixed comments from Ferruh.
---
doc/guides/rel_notes/release_22_03.rst| 7 +++++++ lib/ethdev/ethdev_driver.h | 23 +++++++++++++++++++++++ lib/ethdev/rte_ethdev.c | 17 +++++++++++++++++ lib/ethdev/rte_ethdev.h | 20 ++++++++++++++++++++ lib/ethdev/version.map | 3 +++
Quoted 5 lines
Quoted 2 lines
Btw, can you please confirm that there will be a PMD implementation in this release, (it can be after -rc1)?
YES, I will send a set of patches about hns3 PMD implementation once the API is accepted.
ack, thanks.
Quoted 4 lines
Quoted 4 lines
Because our policy requires at least one PMD implementation is implemented to accept a new API.
(Thomas/David, please correct me if I remember the policy wrong) .
Re: [PATCH v3] ethdev: introduce dump API
Quoted 28 lines · Ferruh Yigit
Quoted 27 lines · Min Hu (Connor)
Hi, Ferruh,
Quoted 24 lines · Ferruh Yigit
Quoted 21 lines · Min Hu (Connor)
Added the ethdev dump API which provides querying private info from ethdev. There exists many private properties in different PMD drivers, such as adapter state, Rx/Tx func algorithm in hns3 PMD. The information of these properties is important for debug. As the information is private, the new API is introduced.
Signed-off-by Min Hu (Connor)<[email protected]>Acked-by Morten Brørup<[email protected]>Acked-by Ray Kinsella<[email protected]>Acked-by Ajit Khaparde<[email protected]>---
v3:
* change 'ethdev' to 'device'
v2:
* fixed comments from Ferruh.
---
doc/guides/rel_notes/release_22_03.rst| 7 +++++++ lib/ethdev/ethdev_driver.h | 23 +++++++++++++++++++++++ lib/ethdev/rte_ethdev.c | 17 +++++++++++++++++ lib/ethdev/rte_ethdev.h | 20 ++++++++++++++++++++ lib/ethdev/version.map | 3 +++
Quoted 8 lines
Quoted 5 lines
Quoted 2 lines
Btw, can you please confirm that there will be a PMD implementation in this release, (it can be after -rc1)?
YES, I will send a set of patches about hns3 PMD implementation once the API is accepted.
ack, thanks.
in fact process document [1] requires at least draft PMD implementation ready to apply the API change [2].
Can be possible to send a draft, simple PMD implementation tomorrow, to justify API design? It can be improved later after -rc1 with new versions.
[1]doc/guides/contributing/patches.rst
[2]
* At least one PMD should implement the API.
It may be a draft sent in a separate series.
Re: [PATCH v3] ethdev: introduce dump API
Hi, Ferruh,
Quoted 35 lines · Ferruh Yigit
Quoted 34 lines · Ferruh Yigit
Quoted 32 lines · Min Hu (Connor)
Hi, Ferruh,
Quoted 24 lines · Ferruh Yigit
Quoted 10 lines · Min Hu (Connor)
Added the ethdev dump API which provides querying private info from ethdev. There exists many private properties in different PMD drivers, such as adapter state, Rx/Tx func algorithm in hns3 PMD. The information of these properties is important for debug. As the information is private, the new API is introduced.
Signed-off-by Min Hu (Connor)<[email protected]>Acked-by Morten Brørup<[email protected]>Acked-by Ray Kinsella<[email protected]>Acked-by Ajit Khaparde<[email protected]>Quoted 11 lines
---
v3:
* change 'ethdev' to 'device'
v2:
* fixed comments from Ferruh.
---
doc/guides/rel_notes/release_22_03.rst| 7 +++++++ lib/ethdev/ethdev_driver.h | 23 +++++++++++++++++++++++ lib/ethdev/rte_ethdev.c | 17 +++++++++++++++++ lib/ethdev/rte_ethdev.h | 20 ++++++++++++++++++++ lib/ethdev/version.map | 3 +++Quoted 2 lines
Btw, can you please confirm that there will be a PMD implementation in this release, (it can be after -rc1)?
YES, I will send a set of patches about hns3 PMD implementation once the API is accepted.
Quoted 9 lines
Quoted 2 lines
ack, thanks.
in fact process document [1] requires at least draft PMD implementation ready to apply the API change [2].
Can be possible to send a draft, simple PMD implementation tomorrow, to justify API design? It can be improved later after -rc1 with new versions.
I have sent a set of patches named 'dump device info', which includes:
a.[patch V4] ethdev: introduce dump API
b. a set of hns3 implementation:
net/hns3: dump device basic info
net/hns3: dump device feature capability
net/hns3: dump device MAC info
net/hns3: dump queue info
net/hns3: dump VLAN configuration info
net/hns3: dump flow director basic info
net/hns3: dump TM configuration info
net/hns3: dump flow control info
Please check it out, thanks.
Quoted 8 lines
[1]
doc/guides/contributing/patches.rst[2]
* At least one PMD should implement the API.
It may be a draft sent in a separate series.
.