Send updated port_id in vdev_init functions
PATCH v2 1/1 Send updated port_id in vdev_init functions
Instead of executing following functions before and after vdev_init
Quoted 2 lines
rte_eth_dev_saveandrte_eth_dev_get_changed_port
update following functions to return allocated port_id.
Quoted 2 lines
rte_eal_vdev_initrte_eal_probe_one
Thanks to Tetsuya for his valuable inputs.
Ravi Kerur (1):
Change rte_eal_vdev_init to update port_id
drivers/net/enic/enic_ethdev.c | 2 +- lib/librte_eal/common/eal_common_dev.c | 13 ++-- lib/librte_eal/common/eal_common_pci.c | 6 +- lib/librte_eal/common/include/rte_dev.h | 36 +++++++++- lib/librte_eal/common/include/rte_pci.h | 4 +- lib/librte_ether/rte_ethdev.c | 122 +++++++++++++++++--------------- lib/librte_ether/rte_ether_version.map | 8 +++ 7 files changed, 125 insertions(+), 66 deletions(-)
Signature · 1 line
1.9.1
PATCH v2 1/1 Change rte_eal_vdev_init to update port_id
v2:
Quoted 3 lines
Remove
rte_pmd_mpipe_devinitchanges
Userte_eal_compare_pci_addrfor address comparison Use dpdk_2.2 in version map file for new functions
v1:
Changes include
Quoted 2 lines
Modify
rte_eal_vdev_initto return allocatedport_idModifyrte_eal_probe_oneto return allocatedport_id
2. Removed following functions
Quoted 2 lines
rte_eth_dev_saveandrte_eth_dev_get_changed_port
3. Added 2 new functions
Quoted 2 lines
rte_eth_dev_get_port_by_namerte_eth_dev_get_port_by_addr
4. Fix return error(ENOMEM) in function rte_pmd_mpipe_devinit
Compiled on Linux for following targets
Quoted 3 lines
x86_64-native-linuxapp-gcc
x86_64-native-linuxapp-clang
x86_x32-native-linuxapp-gcc
Compiled on FreeBSD for following targets
Quoted 2 lines
x86_64-native-bsdapp-clang
x86_64-native-bsdapp-gcc
Tested on Linux/FreeBSD:
Quoted 10 lines
port attach
eth_null
port start all
port stop all
port close all
port detach 0
port attacheth_null
port start all
port stop all
port close all
port detach 0
Successful run of checkpatch.pl on the diffs
Successful validate_abi on Linux for following targets
Quoted 2 lines
x86_64-native-linuxapp-gcc
x86_64-native-linuxapp-clang
--- drivers/net/enic/enic_ethdev.c | 2 +- lib/librte_eal/common/eal_common_dev.c | 13 ++-- lib/librte_eal/common/eal_common_pci.c | 6 +- lib/librte_eal/common/include/rte_dev.h | 36 +++++++++- lib/librte_eal/common/include/rte_pci.h | 4 +- lib/librte_ether/rte_ethdev.c | 122 +++++++++++++++++--------------- lib/librte_ether/rte_ether_version.map | 8 +++ 7 files changed, 125 insertions(+), 66 deletions(-)
drivers/net/enic/enic_ethdev.c +1 -1
@@ -36,8 +36,8 @@#include <stdio.h> #include <stdint.h> -#include <rte_dev.h> #include <rte_pci.h> +#include <rte_dev.h> #include <rte_ethdev.h> #include <rte_string_fns.h>
lib/librte_eal/common/eal_common_dev.c +9 -4
@@ -37,6 +37,7 @@#include <inttypes.h> #include <sys/queue.h> +#include <rte_pci.h> #include <rte_dev.h> #include <rte_devargs.h> #include <rte_debug.h>
@@ -64,7 +65,7 @@ rte_eal_driver_unregister(struct rte_driver *driver) }int -rte_eal_vdev_init (const char *name,const char *args) +rte_eal_vdev_init (const char *name,const char *args,uint8_t *port_id) {struct rte_driver *driver;
@@ -81,8 +82,12 @@ rte_eal_vdev_init(const char *name, const char *args)* will be "eth_pcap", but "name" will be "eth_pcapN". * So use strncmp to compare. */ -if (!strncmp (driver->name, name,strlen (driver->name))) -return driver->init (name, args); +if (!strncmp (driver->name, name,strlen (driver->name))) { +if (!driver->init (name, args)) +return rte_eth_dev_get_port_by_name ( + name, port_id); + } + }RTE_LOG (ERR ,EAL ,"no driver found for %s\n" , name);
@@ -108,7 +113,7 @@ rte_eal_dev_init(void)continue ;if (rte_eal_vdev_init (devargs->virtual.drv_name, - devargs->args)) { + devargs->args,NULL )) {RTE_LOG (ERR ,EAL ,"failed to initialize %s device\n" , devargs->virtual.drv_name);return -1 ;
lib/librte_eal/common/eal_common_pci.c +4 -2
@@ -79,6 +79,7 @@#include <rte_string_fns.h> #include <rte_common.h> #include <rte_devargs.h> +#include <rte_dev.h> #include "eal_private.h"
@@ -322,7 +323,7 @@ pci_detach_all_drivers(struct rte_pci_device *dev)* the driver of the devive. */ int -rte_eal_pci_probe_one (const struct rte_pci_addr *addr) +rte_eal_pci_probe_one (const struct rte_pci_addr *addr,uint8_t *port_id) {struct rte_pci_device *dev =NULL ;int ret =0 ;
@@ -337,7 +338,8 @@ rte_eal_pci_probe_one(const struct rte_pci_addr *addr) ret =pci_probe_all_drivers (dev);if (ret <0 ) goto err_return; -return 0 ; + +return rte_eth_dev_get_port_by_addr (addr, port_id); }return -1 ;
lib/librte_eal/common/include/rte_dev.h +35 -1
@@ -110,10 +110,12 @@ int rte_eal_dev_init(void);* The pointer to a driver name to be initialized. * @param args * The pointer to arguments used by driver initialization. +* @param port_id +* The port_id associated with the driver after initialization. * @return * 0 on success, negative on error */ -int rte_eal_vdev_init (const char *name,const char *args); +int rte_eal_vdev_init (const char *name,const char *args,uint8_t *port_id);/** * Uninitalize a driver specified by name.
@@ -125,6 +127,38 @@ int rte_eal_vdev_init(const char *name, const char *args);*/ int rte_eal_vdev_uninit (const char *name); +/** +* Given name, return port_id associated with the device. +* +* @param name +* Name associated with device. +* @param port_id +* The port identifier of the device. +* +* @return +* - 0: Success. +* - -EINVAL: NULL string (name) +* - -ENODEV failure +*/ +extern int +rte_eth_dev_get_port_by_name (const char *name,uint8_t *port_id); + +/** +* Given pci_addr, return port_id associated with the device. +* +* @param addr +* PCI addr associated with device. +* @param port_id +* The port identifier of the device. +* +* @return +* - 0: Success. +* - -EINVAL: NULL pointer +* - -ENODEV failure +*/ +extern int +rte_eth_dev_get_port_by_addr (const struct rte_pci_addr *addr,uint8_t *port_id); +#define PMD_REGISTER_DRIVER (d)\ void devinitfn_ ##d(void);\ void __attribute__((constructor, used)) devinitfn_ ##d(void)\
lib/librte_eal/common/include/rte_pci.h +3 -1
@@ -406,11 +406,13 @@ void pci_unmap_resource(void *requested_addr, size_t size);* * @param addr * The PCI Bus-Device-Function address to probe. +* @param port_id +* Port-id associated with PCI device address * @return * - 0 on success. * - Negative on error. */ -int rte_eal_pci_probe_one (const struct rte_pci_addr *addr); +int rte_eal_pci_probe_one (const struct rte_pci_addr *addr,uint8_t *port_id);/** * Close the single PCI device.
lib/librte_ether/rte_ethdev.c +65 -57
160 lines not on this page · raw message
lib/librte_ether/rte_ether_version.map +8 -0
@@ -127,3 +127,11 @@ DPDK_2.1 { rte_eth_timesync_read_tx_timestamp; } DPDK_2.0; + +DPDK_2.2 { + global: + + rte_eth_dev_get_port_by_name; + rte_eth_dev_get_port_by_addr; + +} DPDK_2.1;
Signature · 1 line
1.9.1
Re: [PATCH v2] Change rte_eal_vdev_init to update port_id
Quoted 96 lines · Ravi Kerur
v2:
Quoted 3 lines
Remove
rte_pmd_mpipe_devinitchanges
Userte_eal_compare_pci_addrfor address comparison Use dpdk_2.2 in version map file for new functionsv1:
Changes includeQuoted 2 lines
Modify
rte_eal_vdev_initto return allocatedport_idModifyrte_eal_probe_oneto return allocatedport_id2. Removed following functions
Quoted 2 lines
rte_eth_dev_saveandrte_eth_dev_get_changed_port3. Added 2 new functions
Quoted 2 lines
rte_eth_dev_get_port_by_namerte_eth_dev_get_port_by_addr4. Fix return error(ENOMEM) in function
rte_pmd_mpipe_devinitCompiled on Linux for following targets
Quoted 3 lines
x86_64-native-linuxapp-gcc
x86_64-native-linuxapp-clang
x86_x32-native-linuxapp-gccCompiled on FreeBSD for following targets
Quoted 2 lines
x86_64-native-bsdapp-clang
x86_64-native-bsdapp-gccTested on Linux/FreeBSD:
Quoted 10 lines
port attach
eth_null
port start all
port stop all
port close all
port detach 0
port attacheth_null
port start all
port stop all
port close all
port detach 0Successful run of checkpatch.pl on the diffs
Successful
validate_abion Linux for following targetsQuoted 2 lines
x86_64-native-linuxapp-gcc
x86_64-native-linuxapp-clangSigned-off-by Ravi Kerur <[email protected]>--- drivers/net/enic/enic_ethdev.c | 2 +- lib/librte_eal/common/eal_common_dev.c | 13 ++-- lib/librte_eal/common/eal_common_pci.c | 6 +- lib/librte_eal/common/include/rte_dev.h | 36 +++++++++- lib/librte_eal/common/include/rte_pci.h | 4 +- lib/librte_ether/rte_ethdev.c | 122 +++++++++++++++++--------------- lib/librte_ether/rte_ether_version.map | 8 +++ 7 files changed, 125 insertions(+), 66 deletions(-)drivers/net/enic/enic_ethdev.c +1
-1@@ -36,8 +36,8 @@#include <stdio.h> #include <stdint.h> -#include <rte_dev.h> #include <rte_pci.h> +#include <rte_dev.h> #include <rte_ethdev.h> #include <rte_string_fns.h> lib/librte_eal/common/eal_common_dev.c +2
-1@@ -37,6 +37,7 @@#include <inttypes.h> #include <sys/queue.h> +#include <rte_pci.h> #include <rte_dev.h> #include <rte_devargs.h> #include <rte_debug.h> @@ -64,7 +65,7 @@ rte_eal_driver_unregister(struct rte_driver *driver) }int -rte_eal_vdev_init (const char *name,const char *args) +rte_eal_vdev_init (const char *name,const char *args,uint8_t *port_id) {struct rte_driver *driver;@@ -81,8 +82,12 @@ rte_eal_vdev_init(const char *name, const char *args)* will be "eth_pcap", but "name" will be "eth_pcapN".
lib/librte_eal/common/eal_common_dev.c rte_eal_vdev_init(const char *name, const char *args) · line 90
Quoted 10 lines
lib/librte_eal/common/eal_common_dev.c +6
-2* So use strncmp to compare. */ -if (!strncmp (driver->name, name,strlen (driver->name))) -return driver->init (name, args); +if (!strncmp (driver->name, name,strlen (driver->name))) { +if (!driver->init (name, args)) +return rte_eth_dev_get_port_by_name ( + name, port_id); + } +
Please remove needless line.
Quoted 67 lines
lib/librte_eal/common/eal_common_dev.c +1
-1}RTE_LOG (ERR ,EAL ,"no driver found for %s\n" , name);@@ -108,7 +113,7 @@ rte_eal_dev_init(void)continue ;if (rte_eal_vdev_init (devargs->virtual.drv_name, - devargs->args)) { + devargs->args,NULL )) {RTE_LOG (ERR ,EAL ,"failed to initialize %s device\n" , devargs->virtual.drv_name);return -1 ;lib/librte_eal/common/eal_common_pci.c +4
-2@@ -79,6 +79,7 @@#include <rte_string_fns.h> #include <rte_common.h> #include <rte_devargs.h> +#include <rte_dev.h> #include "eal_private.h" @@ -322,7 +323,7 @@ pci_detach_all_drivers(struct rte_pci_device *dev)* the driver of the devive. */ int -rte_eal_pci_probe_one (const struct rte_pci_addr *addr) +rte_eal_pci_probe_one (const struct rte_pci_addr *addr,uint8_t *port_id) {struct rte_pci_device *dev =NULL ;int ret =0 ;@@ -337,7 +338,8 @@ rte_eal_pci_probe_one(const struct rte_pci_addr *addr) ret =pci_probe_all_drivers (dev);if (ret <0 ) goto err_return; -return 0 ; + +return rte_eth_dev_get_port_by_addr (addr, port_id); }return -1 ;lib/librte_eal/common/include/rte_dev.h +5
-1@@ -110,10 +110,12 @@ int rte_eal_dev_init(void);* The pointer to a driver name to be initialized. * @param args * The pointer to arguments used by driver initialization. +* @param port_id +* The port_id associated with the driver after initialization. * @return * 0 on success, negative on error */ -int rte_eal_vdev_init (const char *name,const char *args); +int rte_eal_vdev_init (const char *name,const char *args,uint8_t *port_id);/** * Uninitalize a driver specified by name. @@ -125,6 +127,38 @@ int rte_eal_vdev_init(const char *name, const char *args);*/ int rte_eal_vdev_uninit (const char *name); +/** +* Given name, return port_id associated with the device.
lib/librte_eal/common/include/rte_dev.h int rte_eal_vdev_init(const char *name, const char *args); · line 141
Quoted 10 lines
lib/librte_eal/common/include/rte_dev.h +10
-0+* +* @param name +* Name associated with device. +* @param port_id +* The port identifier of the device. +* +* @return +* - 0: Success. +* - -EINVAL: NULL string (name) +* - -ENODEV failure
Please define above in 'rte_ethdev.h.'
Also Above description uses different format.
The other description in DPDK headers is like below.
"0 on success, negative on error"
So probably it'll be nice to follow it.
At least, it will be nice to choose whether ':' is involved or not.
Quoted 6 lines
lib/librte_eal/common/include/rte_dev.h +6
-0+*/ +extern int +rte_eth_dev_get_port_by_name (const char *name,uint8_t *port_id); + +/** +* Given pci_addr, return port_id associated with the device.
lib/librte_eal/common/include/rte_dev.h int rte_eal_vdev_init(const char *name, const char *args); · line 157
Quoted 10 lines
lib/librte_eal/common/include/rte_dev.h +10
-0+* +* @param addr +* PCI addr associated with device. +* @param port_id +* The port identifier of the device. +* +* @return +* - 0: Success. +* - -EINVAL: NULL pointer +* - -ENODEV failure
Same above.
(Please define it in 'rte_ethdev.h', also use same style)
Quoted 200 lines
lib/librte_eal/common/include/rte_dev.h +4
-0+*/ +extern int +rte_eth_dev_get_port_by_addr (const struct rte_pci_addr *addr,uint8_t *port_id); +#define PMD_REGISTER_DRIVER (d)\ void devinitfn_ ##d(void);\ void __attribute__((constructor, used)) devinitfn_ ##d(void)\lib/librte_eal/common/include/rte_pci.h +3
-1@@ -406,11 +406,13 @@ void pci_unmap_resource(void *requested_addr, size_t size);* * @param addr * The PCI Bus-Device-Function address to probe. +* @param port_id +* Port-id associated with PCI device address * @return * - 0 on success. * - Negative on error. */ -int rte_eal_pci_probe_one (const struct rte_pci_addr *addr); +int rte_eal_pci_probe_one (const struct rte_pci_addr *addr,uint8_t *port_id);/** * Close the single PCI device. lib/librte_ether/rte_ethdev.c +65
-57@@ -442,32 +442,6 @@ rte_eth_dev_get_device_type(uint8_t port_id) }static int -rte_eth_dev_save (struct rte_eth_dev *devs,size_t size) -{ -if ((devs ==NULL ) || - (size !=sizeof (struct rte_eth_dev ) *RTE_MAX_ETHPORTS )) -return -EINVAL ; - -/* save current rte_eth_devices */ -memcpy (devs, rte_eth_devices, size); -return 0 ; -} - -static int -rte_eth_dev_get_changed_port (struct rte_eth_dev *devs,uint8_t *port_id) -{ -if ((devs ==NULL ) || (port_id ==NULL )) -return -EINVAL ; - -/* check which port was attached or detached */ -for (*port_id =0 ; *port_id <RTE_MAX_ETHPORTS ; (*port_id)++, devs++) { -if (rte_eth_devices[*port_id].attached ^ devs->attached) -return 0 ; - } -return -ENODEV ; -} - -static int rte_eth_dev_get_addr_by_port (uint8_t port_id,struct rte_pci_addr *addr) {VALID_PORTID_OR_ERR_RET (port_id, -EINVAL );@@ -530,30 +504,16 @@ rte_eth_dev_is_detachable(uint8_t port_id)static int rte_eth_dev_attach_pdev (struct rte_pci_addr *addr,uint8_t *port_id) { -uint8_t new_port_id; -struct rte_eth_dev devs[RTE_MAX_ETHPORTS ]; -if ((addr ==NULL ) || (port_id ==NULL )) goto err; -/* save current port status */ -if (rte_eth_dev_save (devs,sizeof (devs))) - goto err;/* re-construct pci_device_list */ if (rte_eal_pci_scan ()) goto err; -/* invoke probe func of the driver can handle the new device. -* TODO: -* rte_eal_pci_probe_one() should return port_id. -* And rte_eth_dev_save() and rte_eth_dev_get_changed_port() -* should be removed. */ -if (rte_eal_pci_probe_one (addr)) - goto err; -/* get port_id enabled by above procedures */ -if (rte_eth_dev_get_changed_port (devs, &new_port_id)) +/* Invoke probe func of the driver can handle the new device. */ +if (rte_eal_pci_probe_one (addr, port_id)) goto err; - *port_id = new_port_id;return 0 ; err:RTE_LOG (ERR ,EAL ,"Driver, cannot attach the device\n" );@@ -600,8 +560,6 @@ static int rte_eth_dev_attach_vdev(const char *vdevargs, uint8_t *port_id) { char *name = NULL, *args = NULL; -uint8_t new_port_id; -struct rte_eth_dev devs[RTE_MAX_ETHPORTS ];int ret =-1 ;if ((vdevargs ==NULL ) || (port_id ==NULL ))@@ -611,22 +569,15 @@ rte_eth_dev_attach_vdev(const char *vdevargs, uint8_t *port_id)if (rte_eal_parse_devargs_str (vdevargs, &name, &args)) goto end; -/* save current port status */ -if (rte_eth_dev_save (devs,sizeof (devs))) - goto end;/* walk around dev_driver_list to find the driver of the device, -* then invoke probe function o the driver. -* TODO: -* rte_eal_vdev_init() should return port_id, -* And rte_eth_dev_save() and rte_eth_dev_get_changed_port() -* should be removed. */ -if (rte_eal_vdev_init (name, args)) - goto end; -/* get port_id enabled by above procedures */ -if (rte_eth_dev_get_changed_port (devs, &new_port_id)) +* then invoke probe function of the driver. +* rte_eal_vdev_init() updates port_id allocated after +* initialization. +*/ +if (rte_eal_vdev_init (name, args, port_id)) goto end; + ret =0 ; - *port_id = new_port_id; end:if (name)free (name);@@ -3610,3 +3561,60 @@ rte_eth_dev_set_eeprom(uint8_t port_id, struct rte_dev_eeprom_info *info)FUNC_PTR_OR_ERR_RET (*dev->dev_ops->set_eeprom, -ENOTSUP );return (*dev->dev_ops->set_eeprom)(dev, info); } + +int +rte_eth_dev_get_port_by_name (const char *name,uint8_t *port_id) +{ +int i; + +if (name ==NULL ) { +PMD_DEBUG_TRACE ("Null pointer is specified\n" ); +return -EINVAL ; + } + +if (port_id) + *port_id =RTE_MAX_ETHPORTS ; + +for (i =0 ; i <RTE_MAX_ETHPORTS ; i++) { + +if (!strncmp (name, + rte_eth_dev_data[i].name,strlen (name))) { + +if (port_id) + *port_id = i; + +return 0 ; + } + } +return -ENODEV ; +} + +int +rte_eth_dev_get_port_by_addr (const struct rte_pci_addr *addr,uint8_t *port_id) +{ +int i; +struct rte_pci_device *pci_dev =NULL ; + +if (addr ==NULL ) { +PMD_DEBUG_TRACE ("Null pointer is specified\n" ); +return -EINVAL ; + } + +if (port_id) + *port_id =RTE_MAX_ETHPORTS ; + +for (i =0 ; i <RTE_MAX_ETHPORTS ; i++) { + + pci_dev = rte_eth_devices[i].pci_dev; + +if (pci_dev && + !rte_eal_compare_pci_addr (&pci_dev->addr, addr)) { + +if (port_id) + *port_id = i; + +return 0 ; + } + } +return -ENODEV ; +}lib/librte_ether/rte_ether_version.map +0
-0
lib/librte_ether/rte_ether_version.map DPDK_2.1 { · line 135
Quoted 9 lines
lib/librte_ether/rte_ether_version.map +6
-0@@ -127,3 +127,11 @@ DPDK_2.1 { rte_eth_timesync_read_tx_timestamp; } DPDK_2.0; + +DPDK_2.2 { + global: + + rte_eth_dev_get_port_by_name; + rte_eth_dev_get_port_by_addr;
Please use alphabetical order.
Quoted 2 lines
lib/librte_ether/rte_ether_version.map +2
-0+ +} DPDK_2.1;
Re: [PATCH v2] Change rte_eal_vdev_init to update port_id
Quoted 197 lines · Tetsuya Mukawa
Quoted 48 lines · Ravi Kerur
v2:
Quoted 3 lines
Remove
rte_pmd_mpipe_devinitchanges
Userte_eal_compare_pci_addrfor address comparison Use dpdk_2.2 in version map file for new functionsv1:
Changes includeQuoted 2 lines
Modify
rte_eal_vdev_initto return allocatedport_idModifyrte_eal_probe_oneto return allocatedport_id2. Removed following functions
Quoted 2 lines
rte_eth_dev_saveandrte_eth_dev_get_changed_port3. Added 2 new functions
Quoted 2 lines
rte_eth_dev_get_port_by_namerte_eth_dev_get_port_by_addr4. Fix return error(ENOMEM) in function
rte_pmd_mpipe_devinitCompiled on Linux for following targets
Quoted 3 lines
x86_64-native-linuxapp-gcc
x86_64-native-linuxapp-clang
x86_x32-native-linuxapp-gccCompiled on FreeBSD for following targets
Quoted 2 lines
x86_64-native-bsdapp-clang
x86_64-native-bsdapp-gccTested on Linux/FreeBSD:
Quoted 10 lines
port attach
eth_null
port start all
port stop all
port close all
port detach 0
port attacheth_null
port start all
port stop all
port close all
port detach 0Successful run of checkpatch.pl on the diffs
Successful
validate_abion Linux for following targetsQuoted 2 lines
x86_64-native-linuxapp-gcc
x86_64-native-linuxapp-clangQuoted 8 lines
Signed-off-by Ravi Kerur <[email protected]>--- drivers/net/enic/enic_ethdev.c | 2 +- lib/librte_eal/common/eal_common_dev.c | 13 ++-- lib/librte_eal/common/eal_common_pci.c | 6 +- lib/librte_eal/common/include/rte_dev.h | 36 +++++++++- lib/librte_eal/common/include/rte_pci.h | 4 +- lib/librte_ether/rte_ethdev.c | 122+++++++++++++++++---------------
Review
Quoted 4 lines
lib/librte_ether/rte_ether_version.map | 8 +++ 7 files changed, 125 insertions(+), 66 deletions(-)(quoted diff) +0
-0
b/drivers/net/enic/enic_ethdev.cReview
Quoted 14 lines
index 8280cea..472ef5a 100644
drivers/net/enic/enic_ethdev.c +1
-1@@ -36,8 +36,8 @@#include <stdio.h> #include <stdint.h> -#include <rte_dev.h> #include <rte_pci.h> +#include <rte_dev.h> #include <rte_ethdev.h> #include <rte_string_fns.h> (quoted diff) +0
-0
b/lib/librte_eal/common/eal_common_dev.cQuoted 22 lines
index 4089d66..ffdb3b5 100644
lib/librte_eal/common/eal_common_dev.c +2
-1@@ -37,6 +37,7 @@#include <inttypes.h> #include <sys/queue.h> +#include <rte_pci.h> #include <rte_dev.h> #include <rte_devargs.h> #include <rte_debug.h> @@ -64,7 +65,7 @@ rte_eal_driver_unregister(struct rte_driver *driver) }int -rte_eal_vdev_init (const char *name,const char *args) +rte_eal_vdev_init (const char *name,const char *args,uint8_t *port_id) {struct rte_driver *driver;@@ -81,8 +82,12 @@ rte_eal_vdev_init(const char *name, const char *args)* will be "eth_pcap", but "name" will be "eth_pcapN". lib/librte_eal/common/eal_common_dev.c rte_eal_vdev_init(const char *name, const char *args) · line 90
Quoted 10 lines
lib/librte_eal/common/eal_common_dev.c +6
-2* So use strncmp to compare. */ -if (!strncmp (driver->name, name,strlen (driver->name))) -return driver->init (name, args); +if (!strncmp (driver->name, name,strlen (driver->name))) { +if (!driver->init (name, args)) +return rte_eth_dev_get_port_by_name ( + name, port_id); + } +Please remove needless line.
lib/librte_eal/common/eal_common_dev.c rte_eal_dev_init(void) · line 117
Quoted 10 lines
lib/librte_eal/common/eal_common_dev.c +1
-1}RTE_LOG (ERR ,EAL ,"no driver found for %s\n" , name);@@ -108,7 +113,7 @@ rte_eal_dev_init(void)continue ;if (rte_eal_vdev_init (devargs->virtual.drv_name, - devargs->args)) { + devargs->args,NULL )) { RTE_LOG(ERR, EAL, "failed to initialize %sdevice\n",
Review
Quoted 3 lines
lib/librte_eal/common/eal_common_dev.c +0
-0devargs->virtual.drv_name);return -1 ;(quoted diff) +0
-0
b/lib/librte_eal/common/eal_common_pci.clib/librte_eal/common/eal_common_pci.c rte_eal_pci_probe_one(const struct rte_pci_addr
Quoted 21 lines
index 16e8629..3d97892 100644
lib/librte_eal/common/eal_common_pci.c +2
-1@@ -79,6 +79,7 @@#include <rte_string_fns.h> #include <rte_common.h> #include <rte_devargs.h> +#include <rte_dev.h> #include "eal_private.h" @@ -322,7 +323,7 @@ pci_detach_all_drivers(struct rte_pci_device *dev)* the driver of the devive. */ int -rte_eal_pci_probe_one (const struct rte_pci_addr *addr) +rte_eal_pci_probe_one (const struct rte_pci_addr *addr,uint8_t *port_id) {struct rte_pci_device *dev =NULL ;int ret =0 ;@@ -337,7 +338,8 @@ rte_eal_pci_probe_one(const struct rte_pci_addr*addr)
Review
Quoted 10 lines
lib/librte_eal/common/eal_common_pci.c +2
-1ret =pci_probe_all_drivers (dev);if (ret <0 ) goto err_return; -return 0 ; + +return rte_eth_dev_get_port_by_addr (addr, port_id); }return -1 ;(quoted diff) +0
-0
b/lib/librte_eal/common/include/rte_dev.hQuoted 4 lines
index f601d21..564cdf3 100644
lib/librte_eal/common/include/rte_dev.h int rte_eal_dev_init(void); · line 118
Quoted 10 lines
lib/librte_eal/common/include/rte_dev.h +3
-1@@ -110,10 +110,12 @@ int rte_eal_dev_init(void);* The pointer to a driver name to be initialized. * @param args * The pointer to arguments used by driver initialization. +* @param port_id +* The port_id associated with the driver after initialization. * @return * 0 on success, negative on error */ -int rte_eal_vdev_init (const char *name,const char *args); +int rte_eal_vdev_init(const char *name, const char *args, uint8_t
*port_id);lib/librte_eal/common/include/rte_dev.h int rte_eal_vdev_init(const char *name, const char
Quoted 4 lines
lib/librte_eal/common/include/rte_dev.h +0
-0/** * Uninitalize a driver specified by name. @@ -125,6 +127,38 @@ int rte_eal_vdev_init(const char *name, const char*args);
Quoted 5 lines
lib/librte_eal/common/include/rte_dev.h +2
-0*/ int rte_eal_vdev_uninit (const char *name); +/** +* Given name, return port_id associated with the device. lib/librte_eal/common/include/rte_dev.h int rte_eal_vdev_init(const char *name, const char *args); · line 141
Quoted 10 lines
lib/librte_eal/common/include/rte_dev.h +10
-0+* +* @param name +* Name associated with device. +* @param port_id +* The port identifier of the device. +* +* @return +* - 0: Success. +* - -EINVAL: NULL string (name) +* - -ENODEV failure
Quoted 2 lines
Please define above in '
rte_ethdev.h.'
Hi Tetsuya,
I would like to take a step back and explain why function declarations are in rte_dev.h and not in rte_ethdev.h
Approach 1:
Initially I thought of modifying driver init routine to return/update port_id as the init routine is the place port_id gets allocated and it would have been clean approach. However, it required changes to all PMD_VDEV driver init routine to modify function signature for the changes which I thought may be an overkill.
Approach 2:
Instead I chose to define 2 functions in librte_ether/rte_ethdev.c and make use of it. In this approach new functions are invoked from librte_eal/common/.c to get port_id. If I had new function declarations in rte_ethdev.h and included that file in librte_eal/common/.c files it creates circular dependancy and compilation fails, hence I took hybrid approach of definitions in librte_ether and declarations in librte_eal.
Please let me know if there is a better approach to take care of your comments. As it stands declarations cannot be moved to rte_ethdev.h for compilation reasons.
Thanks,
Ravi
Quoted 254 lines
Also Above description uses different format.
The other description in DPDK headers is like below.
"0 on success, negative on error"
So probably it'll be nice to follow it.
At least, it will be nice to choose whether ':' is involved or not.Quoted 6 lines
lib/librte_eal/common/include/rte_dev.h +6
-0+*/ +extern int +rte_eth_dev_get_port_by_name (const char *name,uint8_t *port_id); + +/** +* Given pci_addr, return port_id associated with the device. lib/librte_eal/common/include/rte_dev.h int rte_eal_vdev_init(const char *name, const char *args); · line 157
Quoted 10 lines
lib/librte_eal/common/include/rte_dev.h +10
-0+* +* @param addr +* PCI addr associated with device. +* @param port_id +* The port identifier of the device. +* +* @return +* - 0: Success. +* - -EINVAL: NULL pointer +* - -ENODEV failure Same above.
(Please define it in 'rte_ethdev.h', also use same style)lib/librte_eal/common/include/rte_dev.h int rte_eal_vdev_init(const char *name, const char *args);
Quoted 3 lines
lib/librte_eal/common/include/rte_dev.h +3
-0+*/ +extern int +rte_eth_dev_get_port_by_addr(const struct rte_pci_addr *addr, uint8_t
*port_id);Review
Quoted 5 lines
lib/librte_eal/common/include/rte_dev.h +1
-0+#define PMD_REGISTER_DRIVER (d)\ void devinitfn_ ##d(void);\ void __attribute__((constructor, used)) devinitfn_ ##d(void)\(quoted diff) +0
-0
b/lib/librte_eal/common/include/rte_pci.hlib/librte_eal/common/include/rte_pci.h void pci_unmap_resource(void *requested_addr,
Quoted 4 lines
index 3fb2d3a..91ad23f 100644
lib/librte_eal/common/include/rte_pci.h +0
-0@@ -406,11 +406,13 @@ void pci_unmap_resource(void *requested_addr,
size_tsize);lib/librte_eal/common/include/rte_pci.h void pci_unmap_resource(void *requested_addr, size_t size);
Quoted 11 lines
lib/librte_eal/common/include/rte_pci.h +3
-1* * @param addr * The PCI Bus-Device-Function address to probe. +* @param port_id +* Port-id associated with PCI device address * @return * - 0 on success. * - Negative on error. */ -int rte_eal_pci_probe_one (const struct rte_pci_addr *addr); +int rte_eal_pci_probe_one(const struct rte_pci_addr *addr, uint8_t
*port_id);Review
Quoted 4 lines
lib/librte_eal/common/include/rte_pci.h +0
-0/** * Close the single PCI device. (quoted diff) +0
-0
b/lib/librte_ether/rte_ethdev.cQuoted 16 lines
index 6b2400c..1b0c6fa 100644
lib/librte_ether/rte_ethdev.c +0
-9@@ -442,32 +442,6 @@ rte_eth_dev_get_device_type(uint8_t port_id) } static int -rte_eth_dev_save (struct rte_eth_dev *devs,size_t size) -{ -if ((devs ==NULL ) || - (size !=sizeof (struct rte_eth_dev ) *RTE_MAX_ETHPORTS )) -return -EINVAL ; - -/* save current rte_eth_devices */ -memcpy (devs, rte_eth_devices, size); -return 0 ;lib/librte_ether/rte_ethdev.c rte_eth_dev_get_device_type(uint8_t port_id) · line 463
Quoted 10 lines
lib/librte_ether/rte_ethdev.c +0
-10-} - -static int -rte_eth_dev_get_changed_port (struct rte_eth_dev *devs,uint8_t *port_id) -{ -if ((devs ==NULL ) || (port_id ==NULL )) -return -EINVAL ; - -/* check which port was attached or detached */ - for (*port_id = 0; *port_id < RTE_MAX_ETHPORTS; (*port_id)++,devs++) {
lib/librte_ether/rte_ethdev.c rte_eth_dev_attach_vdev(const char *vdevargs,
Quoted 53 lines
lib/librte_ether/rte_ethdev.c +2
-25-if (rte_eth_devices[*port_id].attached ^ devs->attached) -return 0 ; - } -return -ENODEV ; -} - -static int rte_eth_dev_get_addr_by_port(uint8_t port_id, struct rte_pci_addr *addr) { VALID_PORTID_OR_ERR_RET(port_id, -EINVAL);@@ -530,30 +504,16 @@ rte_eth_dev_is_detachable(uint8_t port_id)static int rte_eth_dev_attach_pdev (struct rte_pci_addr *addr,uint8_t *port_id) { -uint8_t new_port_id; -struct rte_eth_dev devs[RTE_MAX_ETHPORTS ]; -if ((addr ==NULL ) || (port_id ==NULL )) goto err; -/* save current port status */ -if (rte_eth_dev_save (devs,sizeof (devs))) - goto err;/* re-construct pci_device_list */ if (rte_eal_pci_scan ()) goto err; -/* invoke probe func of the driver can handle the new device. -* TODO: -* rte_eal_pci_probe_one() should return port_id. -* And rte_eth_dev_save() and rte_eth_dev_get_changed_port() -* should be removed. */ -if (rte_eal_pci_probe_one (addr)) - goto err; -/* get port_id enabled by above procedures */ -if (rte_eth_dev_get_changed_port (devs, &new_port_id)) +/* Invoke probe func of the driver can handle the new device. */ +if (rte_eal_pci_probe_one (addr, port_id)) goto err; - *port_id = new_port_id;return 0 ; err:RTE_LOG (ERR ,EAL ,"Driver, cannot attach the device\n" );@@ -600,8 +560,6 @@ static int rte_eth_dev_attach_vdev(const char *vdevargs, uint8_t *port_id) { char *name = NULL, *args = NULL; -uint8_t new_port_id; -struct rte_eth_dev devs[RTE_MAX_ETHPORTS ];int ret =-1 ;if ((vdevargs ==NULL ) || (port_id ==NULL ))@@ -611,22 +569,15 @@ rte_eth_dev_attach_vdev(const char *vdevargs,
uint8_t*port_id)lib/librte_ether/rte_ethdev.c rte_eth_dev_set_eeprom(uint8_t port_id, struct
Quoted 29 lines
lib/librte_ether/rte_ethdev.c +6
-13if (rte_eal_parse_devargs_str (vdevargs, &name, &args)) goto end; -/* save current port status */ -if (rte_eth_dev_save (devs,sizeof (devs))) - goto end;/* walk around dev_driver_list to find the driver of the device, -* then invoke probe function o the driver. -* TODO: -* rte_eal_vdev_init() should return port_id, -* And rte_eth_dev_save() and rte_eth_dev_get_changed_port() -* should be removed. */ -if (rte_eal_vdev_init (name, args)) - goto end; -/* get port_id enabled by above procedures */ -if (rte_eth_dev_get_changed_port (devs, &new_port_id)) +* then invoke probe function of the driver. +* rte_eal_vdev_init() updates port_id allocated after +* initialization. +*/ +if (rte_eal_vdev_init (name, args, port_id)) goto end; + ret =0 ; - *port_id = new_port_id; end:if (name)free (name);@@ -3610,3 +3561,60 @@ rte_eth_dev_set_eeprom(uint8_t port_id, struct
rte_dev_eeprom_info*info)Quoted 23 lines
lib/librte_ether/rte_ethdev.c +20
-0FUNC_PTR_OR_ERR_RET (*dev->dev_ops->set_eeprom, -ENOTSUP );return (*dev->dev_ops->set_eeprom)(dev, info); } + +int +rte_eth_dev_get_port_by_name (const char *name,uint8_t *port_id) +{ +int i; + +if (name ==NULL ) { +PMD_DEBUG_TRACE ("Null pointer is specified\n" ); +return -EINVAL ; + } + +if (port_id) + *port_id =RTE_MAX_ETHPORTS ; + +for (i =0 ; i <RTE_MAX_ETHPORTS ; i++) { + +if (!strncmp (name, + rte_eth_dev_data[i].name,strlen (name))) { + +if (port_id)lib/librte_ether/rte_ethdev.c rte_eth_dev_set_eeprom(uint8_t port_id, struct rte_dev_eeprom_info *info)
Quoted 10 lines
lib/librte_ether/rte_ethdev.c +10
-0+ *port_id = i; + +return 0 ; + } + } +return -ENODEV ; +} + +int +rte_eth_dev_get_port_by_addr(const struct rte_pci_addr *addr, uint8_t
*port_id)Review
Quoted 28 lines
lib/librte_ether/rte_ethdev.c +27
-0+{ +int i; +struct rte_pci_device *pci_dev =NULL ; + +if (addr ==NULL ) { +PMD_DEBUG_TRACE ("Null pointer is specified\n" ); +return -EINVAL ; + } + +if (port_id) + *port_id =RTE_MAX_ETHPORTS ; + +for (i =0 ; i <RTE_MAX_ETHPORTS ; i++) { + + pci_dev = rte_eth_devices[i].pci_dev; + +if (pci_dev && + !rte_eal_compare_pci_addr (&pci_dev->addr, addr)) { + +if (port_id) + *port_id = i; + +return 0 ; + } + } +return -ENODEV ; +}(quoted diff) +0
-0b/lib/librte_ether/rte_ether_version.map
lib/librte_ether/rte_ether_version.map DPDK_2.1 { · line 135
Quoted 13 lines
index 8345a6c..9ab14da 100644
lib/librte_ether/rte_ether_version.map +6
-0@@ -127,3 +127,11 @@ DPDK_2.1 { rte_eth_timesync_read_tx_timestamp; } DPDK_2.0; + +DPDK_2.2 { + global: + + rte_eth_dev_get_port_by_name; + rte_eth_dev_get_port_by_addr;Please use alphabetical order.
Quoted 2 lines
lib/librte_ether/rte_ether_version.map +2
-0+ +} DPDK_2.1;
Re: [PATCH v2] Change rte_eal_vdev_init to update port_id
Quoted 45 lines · Ravi Kerur
Review int rte_eal_vdev_init(const char *name,
Quoted 4 lines
/**
* Uninitalize a driver specified by name.(quoted diff) +0
-0@@ -125,6 +127,38 @@ int rte_eal_vdev_init(const char *name,const char *args); > */ > int rte_eal_vdev_uninit(const char *name); > > +/** > + * Given name, return port_id associated with the device. > + * > + * @param name > + * Name associated with device. > + * @param port_id > + * The port identifier of the device. > + * > + * @return > + * - 0: Success. > + * - -EINVAL: NULL string (name) > + * - -ENODEV failure Please define above in 'rte_ethdev.h.'Hi Tetsuya,
I would like to take a step back and explain why function declarations are in
rte_dev.hand not inrte_ethdev.hApproach 1:
Initially I thought of modifying driver init routine to return/updateport_idas the init routine is the placeport_idgets allocated and it would have been clean approach. However, it required changes to allPMD_VDEVdriver init routine to modify function signature for the changes which I thought may be an overkill.Approach 2:
Instead I chose to define 2 functions inlibrte_ether/rte_ethdev.cand make use of it. In this approach new functions are invoked from librte_eal/common/.c to getport_id. If I had new function declarations inrte_ethdev.hand included that file in librte_eal/common/.c files it creates circular dependancy and compilation fails, hence I took hybrid approach of definitions inlibrte_etherand declarations inlibrte_eal.
Quoted 7 lines
Please let me know if there is a better approach to take care of your comments. As it stands declarations cannot be moved to
rte_ethdev.hfor compilation reasons.Thanks,
Ravi
Hi Ravi,
(Adding David)
I appreciate your description. I understand why you define the functions in rte_dev.h.
About Approach2, I don't know a way to implement cleanly.
I guess if we define the functions in rte_dev.h, the developers who want to use the functions will be confused because the functions are implemented in ethdev.c, but it is needed to include rte_dev.h.
To avoid such a confusion, following implementation might be worked, but I am not sure this cording style is allowed in eal library.
----------------------------
Define the functions in rte_ethdev.h, then fix librte_eal/common/.c files like below
ex) lib/librte_eal/common/eal_common_dev.c
----------------------------
+#include <rte_pci.h>
#include <rte_dev.h>
#include <rte_devargs.h>
#include <rte_debug.h>
#include "eal_private.h"
+extern int rte_eth_dev_get_port_by_name(const char *name, uint8_t *port_id);
+extern int rte_eth_dev_get_port_by_addr(const struct rte_pci_addr *addr, uint8_t *port_id);
----------------------------
In this case, the developer might be able to notice that above usage in eal library is some kind of exception. But I guess the DPDK code won't be clean if we start having a exception.
So it might be good to choose Approach1, because apparently it is straight forward.
Anyone won't be confused and complained about coding style.
Hi David,
Could you please let us know what you think?
Do you have a good approach for this?
Thanks,
Tetsuya
Re: [PATCH v2] Change rte_eal_vdev_init to update port_id
Hi Thomas, David
Let us know how you want us to fix this? To fix rte_eal_vdev_init and rte_eal_pci_probe_one to return allocated port_id we had 2 approaches mentioned in earlier discussion. In addition to those we have another approach with changes isolated only to rte_ether component. I am attaching diffs (preliminary) with this email. Please let us know your inputs since it involves EAL component.
Thanks,
Ravi
Quoted 104 lines · Tetsuya Mukawa
Quoted 45 lines · Ravi Kerur
Review int rte_eal_vdev_init(const char *name,
Quoted 4 lines
/**
* Uninitalize a driver specified by name.(quoted diff) +0
-0@@ -125,6 +127,38 @@ int rte_eal_vdev_init(const char *name,const char *args); > */ > int rte_eal_vdev_uninit(const char *name); > > +/** > + * Given name, return port_id associated with the device. > + * > + * @param name > + * Name associated with device. > + * @param port_id > + * The port identifier of the device. > + * > + * @return > + * - 0: Success. > + * - -EINVAL: NULL string (name) > + * - -ENODEV failure Please define above in 'rte_ethdev.h.'Hi Tetsuya,
I would like to take a step back and explain why function declarations are in
rte_dev.hand not inrte_ethdev.hApproach 1:
Initially I thought of modifying driver init routine to return/updateport_idas the init routine is the placeport_idgets allocated and it would have been clean approach. However, it required changes to allPMD_VDEVdriver init routine to modify function signature for the changes which I thought may be an overkill.Approach 2:
Instead I chose to define 2 functions inlibrte_ether/rte_ethdev.cand make use of it. In this approach new functions are invoked from librte_eal/common/.c to getport_id. If I had new function declarations inrte_ethdev.hand included that file in librte_eal/common/.c files it creates circular dependancy and compilation fails, hence I took hybrid approach of definitions inlibrte_etherand declarations inlibrte_eal.Quoted 7 lines
Please let me know if there is a better approach to take care of your comments. As it stands declarations cannot be moved to
rte_ethdev.hfor compilation reasons.Thanks,
RaviHi Ravi,
(Adding David)I appreciate your description. I understand why you define the functions in
rte_dev.h.About Approach2, I don't know a way to implement cleanly.
I guess if we define the functions inrte_dev.h, the developers who want to use the functions will be confused because the functions are implemented inethdev.c, but it is needed to includerte_dev.h.To avoid such a confusion, following implementation might be worked, but I am not sure this cording style is allowed in eal library.
----------------------------
Define the functions inrte_ethdev.h, then fix librte_eal/common/.c files like belowex)
lib/librte_eal/common/eal_common_dev.c
----------------------------lib/librte_eal/common/eal_common_dev.c +2
-0+#include <rte_pci.h> #include <rte_dev.h> #include <rte_devargs.h> #include <rte_debug.h> #include "eal_private.h" +extern int rte_eth_dev_get_port_by_name(const char *name, uint8_t
*port_id);
+extern intrte_eth_dev_get_port_by_addr(const struct rte_pci_addr *addr, uint8_t *port_id);
----------------------------In this case, the developer might be able to notice that above usage in eal library is some kind of exception. But I guess the DPDK code won't be clean if we start having a exception.
So it might be good to choose Approach1, because apparently it is straight forward.
Anyone won't be confused and complained about coding style.Hi David,
Could you please let us know what you think?
Do you have a good approach for this?Thanks,
Tetsuya
Re: [PATCH v2] Change rte_eal_vdev_init to update port_id
Hello Ravi, Tetsuya,
Quoted 7 lines · Ravi Kerur
Let us know how you want us to fix this? To fix
rte_eal_vdev_initandrte_eal_pci_probe_oneto return allocatedport_idwe had 2 approaches mentioned in earlier discussion. In addition to those we have another approach with changes isolated only torte_ethercomponent. I am attaching diffs (preliminary) with this email. Please let us know your inputs since it involves EAL component.
- This patch looks like a good ethdev cleanup (even if it really lacks some context / commit log).
I wonder just why you only take the first part of the name in rte_eth_dev_get_port_by_name().
Would not this match, let's say, both toto and toto0 vdevs ? Is this intended ?
- In the end, with this patch, do we still need to update eal ? Looking at the code, I am not sure anymore.
Signature · 1 line
David Marchand
Re: [PATCH v2] Change rte_eal_vdev_init to update port_id
Hi David,
Quoted 20 lines · David Marchand
Hello Ravi, Tetsuya,
Quoted 7 lines · Ravi Kerur
Let us know how you want us to fix this? To fix
rte_eal_vdev_initandrte_eal_pci_probe_oneto return allocatedport_idwe had 2 approaches mentioned in earlier discussion. In addition to those we have another approach with changes isolated only torte_ethercomponent. I am attaching diffs (preliminary) with this email. Please let us know your inputs since it involves EAL component.- This patch looks like a good ethdev cleanup (even if it really lacks some context / commit log).
I wonder just why you only take the first part of the name in
rte_eth_dev_get_port_by_name().
Would not this match, let's say, both toto and toto0 vdevs ? Is this intended ?
Quoted 6 lines
It was not intended, i will look into it.
- In the end, with this patch, do we still need to update eal ? Looking at the code, I am not sure anymore.
Approach 3 (preliminary diffs sent as an attachment) doesn't involve EAL but the other two solutions do. So please let us know which one you prefer. I will send updated patch.
Thanks,
Ravi
Quoted 6 lines
--
David Marchand
Re: [PATCH v2] Change rte_eal_vdev_init to update port_id
Hi David, Tetsuya,
I have sent V3 (changes isolated to rte_ether component) for formal review. Please look into it and let me know your inputs.
@David: I looked at "rte_eth_dev_get_port_by_name()", this function is similar to "rte_eth_dev_get_name_by_port" and I have used same logic. Let me know if this not correct I can fix both.
Thanks,
Ravi
Quoted 49 lines · Ravi Kerur
Hi David,
Quoted 20 lines · David Marchand
Hello Ravi, Tetsuya,
Quoted 7 lines · Ravi Kerur
Let us know how you want us to fix this? To fix
rte_eal_vdev_initandrte_eal_pci_probe_oneto return allocatedport_idwe had 2 approaches mentioned in earlier discussion. In addition to those we have another approach with changes isolated only torte_ethercomponent. I am attaching diffs (preliminary) with this email. Please let us know your inputs since it involves EAL component.- This patch looks like a good ethdev cleanup (even if it really lacks some context / commit log).
I wonder just why you only take the first part of the name in
rte_eth_dev_get_port_by_name().
Would not this match, let's say, both toto and toto0 vdevs ? Is this intended ?Quoted 6 lines
It was not intended, i will look into it.
- In the end, with this patch, do we still need to update eal ? Looking at the code, I am not sure anymore.
Approach 3 (preliminary diffs sent as an attachment) doesn't involve EAL but the other two solutions do. So please let us know which one you prefer. I will send updated patch.
Thanks,
RaviQuoted 6 lines
--
David Marchand
Re: [PATCH v2] Change rte_eal_vdev_init to update port_id
Hi Ravi,
I am out of office now, and will be back 29th.
After backing, I will check it and send reply.
Tetsuya
Quoted 58 lines · Ravi Kerur
Hi David, Tetsuya,
I have sent V3 (changes isolated to
rte_ethercomponent) for formal review. Please look into it and let me know your inputs.@David: I looked at "
rte_eth_dev_get_port_by_name()", this function is similar to "rte_eth_dev_get_name_by_port" and I have used same logic. Let me know if this not correct I can fix both.Thanks,
RaviQuoted 43 lines · Ravi Kerur
Hi David,
Quoted 14 lines · David Marchand
Hello Ravi, Tetsuya,
Quoted 7 lines · Ravi Kerur
Let us know how you want us to fix this? To fix
rte_eal_vdev_initandrte_eal_pci_probe_oneto return allocatedport_idwe had 2 approaches mentioned in earlier discussion. In addition to those we have another approach with changes isolated only torte_ethercomponent. I am attaching diffs (preliminary) with this email. Please let us know your inputs since it involves EAL component.- This patch looks like a good ethdev cleanup (even if it really lacks some context / commit log).
Quoted 9 lines
I wonder just why you only take the first part of the name in
rte_eth_dev_get_port_by_name().
Would not this match, let's say, both toto and toto0 vdevs ? Is this intended ?It was not intended, i will look into it.
- In the end, with this patch, do we still need to update eal ? Looking at the code, I am not sure anymore.Approach 3 (preliminary diffs sent as an attachment) doesn't involve EAL but the other two solutions do. So please let us know which one you prefer. I will send updated patch.
Thanks,
RaviQuoted 5 lines
--
David Marchand
Re: [PATCH v2] Change rte_eal_vdev_init to update port_id
In reply to Ravi KerurQuoted 4 lines · Ravi Kerur
Hi David, Tetsuya,
I have sent V3 (changes isolated to
rte_ethercomponent) for formal review. Please look into it and let me know your inputs.
Hi Ravi,
I've checked the patch.
I guess your patch is good.
Quoted 4 lines
@David: I looked at "
rte_eth_dev_get_port_by_name()", this function is similar to "rte_eth_dev_get_name_by_port" and I have used same logic. Let me know if this not correct I can fix both.
Do you comment about rte_eth_dev_get_port_by_name and rte_eth_dev_get_port_by_addr?
If so, I guess we don't need to merge.
Quoted 49 lines
Thanks,
RaviQuoted 43 lines · Ravi Kerur
Hi David,
Quoted 14 lines · David Marchand
Hello Ravi, Tetsuya,
Quoted 7 lines · Ravi Kerur
Let us know how you want us to fix this? To fix
rte_eal_vdev_initandrte_eal_pci_probe_oneto return allocatedport_idwe had 2 approaches mentioned in earlier discussion. In addition to those we have another approach with changes isolated only torte_ethercomponent. I am attaching diffs (preliminary) with this email. Please let us know your inputs since it involves EAL component.- This patch looks like a good ethdev cleanup (even if it really lacks some context / commit log).
Quoted 9 lines
I wonder just why you only take the first part of the name in
rte_eth_dev_get_port_by_name().
Would not this match, let's say, both toto and toto0 vdevs ? Is this intended ?It was not intended, i will look into it.
- In the end, with this patch, do we still need to update eal ? Looking at the code, I am not sure anymore.Approach 3 (preliminary diffs sent as an attachment) doesn't involve EAL but the other two solutions do. So please let us know which one you prefer. I will send updated patch.
Thanks,
RaviQuoted 5 lines
--
David Marchand
Re: [PATCH v2] Change rte_eal_vdev_init to update port_id
Hi Tetsuya,
Quoted 13 lines · Tetsuya Mukawa
Quoted 3 lines · Ravi Kerur
Hi David, Tetsuya,
I have sent V3 (changes isolated to
rte_ethercomponent) for formalreview.
Quoted 1 line
Please look into it and let me know your inputs.
Hi Ravi,
I've checked the patch.
I guess your patch is good.Quoted 1 line
Quoted 9 lines
Quoted 3 lines
@David: I looked at "
rte_eth_dev_get_port_by_name()", this function is similar to "rte_eth_dev_get_name_by_port" and I have used same logic. Let me know if this not correct I can fix both.Do you comment about
rte_eth_dev_get_port_by_nameandrte_eth_dev_get_port_by_addr?
If so, I guess we don't need to merge.
I just mentioned that new functions are using same logic as existing function.
Thanks,
Ravi
Quoted 55 lines
Quoted 10 lines
Thanks,
RaviQuoted 4 lines · Ravi Kerur
Hi David,
On Thu, Sep 3, 2015 at 7:04 AM, David Marchand <
Quoted 10 lines
Quoted 10 lines
Quoted 8 lines · wrote:
Hello Ravi, Tetsuya,
Quoted 4 lines · Ravi Kerur
Let us know how you want us to fix this? To fix
rte_eal_vdev_initandrte_eal_pci_probe_oneto return allocatedport_idwe had 2 approaches mentioned in earlier discussion. In addition to those we have another approach with changes isolated only torte_ethercomponent. I amattaching
Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line
diffs (preliminary) with this email. Please let us know your inputs
since
Quoted 10 lines
Quoted 10 lines
Quoted 10 lines
Quoted 2 lines
it involves EAL component.
- This patch looks like a good ethdev cleanup (even if it really lacks some context / commit log).
I wonder just why you only take the first part of the name in
rte_eth_dev_get_port_by_name().
Would not this match, let's say, both toto and toto0 vdevs ? Is this intended ?Quoted 6 lines
Quoted 6 lines
Quoted 4 lines
It was not intended, i will look into it.
- In the end, with this patch, do we still need to update eal ? Looking at the code, I am not sure anymore.Approach 3 (preliminary diffs sent as an attachment) doesn't involve EAL but the other two solutions do. So please let us know which one you
prefer.
Quoted 12 lines
Quoted 12 lines
I will send updated patch.
Thanks,
RaviQuoted 5 lines
--
David Marchand