usb: gadget: udc: Fix duplicate uclass name
PATCH 1/1 usb: gadget: udc: Fix duplicate uclass name
Currently both USB host uclass and USB gadget uclass are using the same name "usb" which break uclass functions like uclass_get_by_name().
Rename the uclass to "usb_gadget" to fix, also makes bind/unbind by class index (or sequence) working.
This breaks the capacity of using "usb" as DT alias sequence numbering which needs a fix afterwards.
--- drivers/usb/gadget/udc/udc-uclass.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/usb/gadget/udc/udc-uclass.c +1 -1
@@ -83,7 +83,7 @@ __weak int dm_usb_gadget_handle_interrupts(struct udevice *dev)#if CONFIG_IS_ENABLED (DM )UCLASS_DRIVER (usb_gadget_generic) = { .id =UCLASS_USB_GADGET_GENERIC , - .name ="usb" , + .name ="usb_gadget" , .flags =DM_UC_FLAG_SEQ_ALIAS , };#endif
--
2.45.2
PATCH 1/1 dm: core: Show device sequence instead in dm_dump_tree()
Currently uclass index is shown in DM tree dump which ignores alias sequence numbering. The result could be confusing since these 2 numbers could be different. Show device sequence number instead as it's more meaningful.
Also update documentation to use sequence number.
--- doc/usage/cmd/dm.rst | 7 +++---- drivers/core/dump.c | 4 ++-- 2 files changed, 5 insertions(+), 6 deletions(-)
doc/usage/cmd/dm.rst +3 -4
@@ -112,9 +112,8 @@ This shows the full tree of devices including the following fields: uclass Shows the name of the uclass for the device -Index - Shows the index number of the device, within the uclass. This shows the - ordering within the uclass, but not the sequence number. +Seq + Shows the sequence number of the device, within the uclass. Probed Shows `+` if the device is active
@@ -366,7 +365,7 @@ dm tree This example shows the abridged sandbox output:: => dm tree - Class Index Probed Driver Name + Class Seq Probed Driver Name ----------------------------------------------------------- root 0 [ + ] root_driver root_driver demo 0 [ ] demo_shape_drv |-- demo_shape_drv
drivers/core/dump.c +2 -2
@@ -40,7 +40,7 @@ static void show_devices(struct udevice *dev, int depth, int last_flag,/* print the first 20 characters to not break the tree-format. */ printf (CONFIG_IS_ENABLED (USE_TINY_PRINTF ) ?" %s %d [ %c ] %s " :" %-10.10s %3d [ %c ] %-20.20s " , dev->uclass->uc_drv->name, -dev_get_uclass_index (dev,NULL ), + dev->seq_, flags &DM_FLAG_ACTIVATED ?'+' :' ' , dev->driver->name);for (i = depth; i >=0 ; i--) {
@@ -129,7 +129,7 @@ void dm_dump_tree(char *dev_name, bool extended, bool sort) {struct udevice *root; -printf (" Class Index Probed Driver Name\n" ); +printf (" Class Seq Probed Driver Name\n" );printf ("-----------------------------------------------------------\n" ); root =dm_root ();
--
2.45.2
Re: [PATCH] dm: core: Show device sequence instead in dm_dump_tree()
Quoted 6 lines · Zixun LI
Currently uclass index is shown in DM tree dump which ignores alias sequence numbering. The result could be confusing since these 2 numbers could be different. Show device sequence number instead as it's more meaningful.
Quoted 7 lines
Also update documentation to use sequence number.
Signed-off-by Zixun LI <[email protected]>--- doc/usage/cmd/dm.rst | 7 +++---- drivers/core/dump.c | 4 ++-- 2 files changed, 5 insertions(+), 6 deletions(-)
Quoted 51 lines
doc/usage/cmd/dm.rst +3
-4@@ -112,9 +112,8 @@ This shows the full tree of devices including the following fields: uclass Shows the name of the uclass for the device -Index - Shows the index number of the device, within the uclass. This shows the - ordering within the uclass, but not the sequence number. +Seq + Shows the sequence number of the device, within the uclass. Probed Shows `+` if the device is active@@ -366,7 +365,7 @@ dm tree This example shows the abridged sandbox output:: => dm tree - Class Index Probed Driver Name + Class Seq Probed Driver Name ----------------------------------------------------------- root 0 [ + ] root_driver root_driver demo 0 [ ] demo_shape_drv |-- demo_shape_drvdrivers/core/dump.c +2
-3@@ -40,7 +40,7 @@ static void show_devices(struct udevice *dev, int depth, int last_flag,/* print the first 20 characters to not break the tree-format. */ printf (CONFIG_IS_ENABLED (USE_TINY_PRINTF ) ?" %s %d [ %c ] %s " :" %-10.10s %3d [ %c ] %-20.20s " , dev->uclass->uc_drv->name, -dev_get_uclass_index (dev,NULL ), + dev->seq_, flags &DM_FLAG_ACTIVATED ?'+' :' ' , dev->driver->name);for (i = depth; i >=0 ; i--) {@@ -129,7 +129,7 @@ void dm_dump_tree(char *dev_name, bool extended, bool sort) {struct udevice *root; -printf (" Class Index Probed Driver Name\n" ); +printf (" Class Seq Probed Driver Name\n" );printf ("-----------------------------------------------------------\n" ); root =dm_root (); --2.45.2
Re: [PATCH] dm: core: Show device sequence instead in dm_dump_tree()
In reply to Zixun LIQuoted 8 lines · Zixun LI
Currently uclass index is shown in DM tree dump which ignores alias sequence numbering. The result could be confusing since these 2 numbers could be different. Show device sequence number instead as it's more meaningful.
Also update documentation to use sequence number.
[...]
Applied to u-boot/next, thanks!
Signature · 1 line
Tom
PATCH 1/1 cmd: bind: Use device sequence instead for driver bind/unbind
In reply to Zixun LICurrently uclass index is used for bind/unbind which ignores alias sequence numbering. Use device sequence number instead as it's the number explicitly set in the DT.
Also update documentation to use sequence number.
--- cmd/bind.c | 46 ++++++++++++++++++++-------------------- doc/usage/cmd/bind.rst | 12 +++++------ doc/usage/cmd/unbind.rst | 14 ++++++------ 3 files changed, 36 insertions(+), 36 deletions(-)
cmd/bind.c +23 -23
@@ -10,8 +10,8 @@#include <dm/root.h> #include <dm/uclass-internal.h> -static int bind_by_class_index (const char *uclass,int index, -const char *drv_name) +static int bind_by_class_seq (const char *uclass,int seq, +const char *drv_name) {static enum uclass_id uclass_id;struct udevice *dev;
@@ -31,9 +31,9 @@ static int bind_by_class_index(const char *uclass, int index,return -EINVAL ; } - ret =uclass_find_device (uclass_id, index, &parent); + ret =uclass_find_device_by_seq (uclass_id, seq, &parent);if (!parent || ret) { -printf ("Cannot find device %d of class %s\n" , index, uclass); +printf ("Cannot find device %d of class %s\n" , seq, uclass);return ret; }
@@ -47,7 +47,7 @@ static int bind_by_class_index(const char *uclass, int index,return 0 ; } -static int find_dev (const char *uclass,int index,struct udevice **devp) +static int find_dev (const char *uclass,int seq,struct udevice **devp) {static enum uclass_id uclass_id;int rc;
@@ -58,21 +58,21 @@ static int find_dev(const char *uclass, int index, struct udevice **devp)return -EINVAL ; } - rc =uclass_find_device (uclass_id, index, devp); + rc =uclass_find_device_by_seq (uclass_id, seq, devp);if (!*devp || rc) { -printf ("Cannot find device %d of class %s\n" , index, uclass); +printf ("Cannot find device %d of class %s\n" , seq, uclass);return rc; }return 0 ; } -static int unbind_by_class_index (const char *uclass,int index) +static int unbind_by_class_seq (const char *uclass,int seq) {int ret;struct udevice *dev; - ret =find_dev (uclass, index, &dev); + ret =find_dev (uclass, seq, &dev);if (ret)return ret;
@@ -91,8 +91,8 @@ static int unbind_by_class_index(const char *uclass, int index)return 0 ; } -static int unbind_child_by_class_index (const char *uclass,int index, -const char *drv_name) +static int unbind_child_by_class_seq (const char *uclass,int seq, +const char *drv_name) {struct udevice *parent;int ret;
@@ -104,7 +104,7 @@ static int unbind_child_by_class_index(const char *uclass, int index,return -ENOENT ; } - ret =find_dev (uclass, index, &parent); + ret =find_dev (uclass, seq, &parent);if (ret)return ret;
@@ -217,19 +217,19 @@ static int do_bind_unbind(struct cmd_tbl *cmdtp, int flag, int argc,return CMD_RET_USAGE ; ret =unbind_by_node_path (argv[1 ]); }else if (!by_node && bind) { -int index = (argc >2 ) ?dectoul (argv[2 ],NULL ) :0 ; +int seq = (argc >2 ) ?dectoul (argv[2 ],NULL ) :0 ;if (argc !=4 )return CMD_RET_USAGE ; - ret =bind_by_class_index (argv[1 ], index, argv[3 ]); + ret =bind_by_class_seq (argv[1 ], seq, argv[3 ]); }else if (!by_node && !bind) { -int index = (argc >2 ) ?dectoul (argv[2 ],NULL ) :0 ; +int seq = (argc >2 ) ?dectoul (argv[2 ],NULL ) :0 ;if (argc ==3 ) - ret =unbind_by_class_index (argv[1 ], index); + ret =unbind_by_class_seq (argv[1 ], seq);else if (argc ==4 ) - ret =unbind_child_by_class_index (argv[1 ], index, - argv[3 ]); + ret =unbind_child_by_class_seq (argv[1 ], seq, + argv[3 ]);else return CMD_RET_USAGE ; }
@@ -244,17 +244,17 @@ U_BOOT_CMD( bind,4 ,0 , do_bind_unbind,"Bind a device to a driver" ,"<node path> <driver>\n" -"bind <class> <index> <driver>\n" +"bind <class> <seq> <driver>\n" "Use 'dm tree' to list all devices registered in the driver model,\n" -"their path, class, index and current driver.\n" +"their path, class, sequence and current driver.\n" );U_BOOT_CMD ( unbind,4 ,0 , do_bind_unbind,"Unbind a device from a driver" ,"<node path>\n" -"unbind <class> <index>\n" -"unbind <class> <index> <driver>\n" +"unbind <class> <seq>\n" +"unbind <class> <seq> <driver>\n" "Use 'dm tree' to list all devices registered in the driver model,\n" -"their path, class, index and current driver.\n" +"their path, class, sequence and current driver.\n" );
doc/usage/cmd/bind.rst +6 -6
@@ -12,7 +12,7 @@ Synopsis :: bind <node path> <driver> - bind <class> <index> <driver> + bind <class> <seq> <driver> Description -----------
@@ -30,8 +30,8 @@ node path class device class name -index - index of the parent device in the device class +seq + sequence number of the parent device in the device class driver device driver name
@@ -46,7 +46,7 @@ using the two alternative bind syntaxes. .. code-block:: => dm tree - Class Index Probed Driver Name + Class Seq Probed Driver Name ----------------------------------------------------------- root 0 [ + ] root_driver root_driver ...
@@ -75,13 +75,13 @@ using the two alternative bind syntaxes. => date Cannot find RTC: err=-19 => dm tree - Class Index Probed Driver Name + Class Seq Probed Driver Name ----------------------------------------------------------- root 0 [ + ] root_driver root_driver ... => bind /pl031@9010000 rtc-pl031 => dm tree - Class Index Probed Driver Name + Class Seq Probed Driver Name ----------------------------------------------------------- root 0 [ + ] root_driver root_driver ...
doc/usage/cmd/unbind.rst +7 -7
@@ -12,8 +12,8 @@ Synopsis :: unbind <node path> - unbind <class> <index> - unbind <class> <index> <driver> + unbind <class> <seq> + unbind <class> <seq> <driver> Description -----------
@@ -27,8 +27,8 @@ node path class device class name -index - index of the device in the device class +seq + sequence number of the device in the device class driver device driver name
@@ -43,7 +43,7 @@ using the three alternative unbind syntaxes. .. code-block:: => dm tree - Class Index Probed Driver Name + Class Seq Probed Driver Name ----------------------------------------------------------- root 0 [ + ] root_driver root_driver ...
@@ -70,7 +70,7 @@ using the three alternative unbind syntaxes. } => unbind /pl031@9010000 => dm tree - Class Index Probed Driver Name + Class Seq Probed Driver Name ----------------------------------------------------------- root 0 [ + ] root_driver root_driver ...
@@ -78,7 +78,7 @@ using the three alternative unbind syntaxes. Cannot find a device with path /pl031@9010000 => bind /pl031@9010000 rtc-pl031 => dm tree - Class Index Probed Driver Name + Class Seq Probed Driver Name ----------------------------------------------------------- root 0 [ + ] root_driver root_driver ...
--
2.45.2
Re: [PATCH] cmd: bind: Use device sequence instead for driver bind/unbind
Quoted 5 lines · Zixun LI
Currently uclass index is used for bind/unbind which ignores alias sequence numbering. Use device sequence number instead as it's the number explicitly set in the DT.
Quoted 9 lines
Also update documentation to use sequence number.
Signed-off-by Zixun LI <[email protected]>--- cmd/bind.c | 46 ++++++++++++++++++++-------------------- doc/usage/cmd/bind.rst | 12 +++++------ doc/usage/cmd/unbind.rst | 14 ++++++------ 3 files changed, 36 insertions(+), 36 deletions(-)
Re: [PATCH] cmd: bind: Use device sequence instead for driver bind/unbind
In reply to Zixun LIHi,
Quoted 8 lines · Zixun LI
Currently uclass index is used for bind/unbind which ignores alias sequence numbering. Use device sequence number instead as it's the number explicitly set in the DT.
Also update documentation to use sequence number.
[...]
Thanks, Applied to https://source.denx.de/u-boot/custodians/u-boot-dfu (u-boot-dfu-next)
[1/1] cmd: bind: Use device sequence instead for driver bind/unbind
https://source.denx.de/u-boot/custodians/u-boot-dfu/-/commit/0c48507eee08b5e7d4e8484727efe47cfabff0ee
--
Mattijs
Re: [PATCH] cmd: bind: Use device sequence instead for driver bind/unbind
Quoted 10 lines · Mattijs Korpershoek
Hi,
Quoted 7 lines · Zixun LI
Currently uclass index is used for bind/unbind which ignores alias sequence numbering. Use device sequence number instead as it's the number explicitly set in the DT.
Also update documentation to use sequence number.
Quoted 6 lines
Quoted 1 line
[...]
Thanks, Applied to https://source.denx.de/u-boot/custodians/u-boot-dfu (u-boot-dfu-next)
[1/1] cmd: bind: Use device sequence instead for driver bind/unbind
https://source.denx.de/u-boot/custodians/u-boot-dfu/-/commit/0c48507eee08b5e7d4e8484727efe47cfabff0ee
Somehow, b4 did the wrong thing here when running:
$ b4 shazam -s -l --check [email protected]
Grabbing thread from lore.kernel.org/all/[email protected]/t.mbox.gz Checking for newer revisions
Grabbing search results from lore.kernel.org
Analyzing 12 messages in the thread
Assuming new revision: v2 ([PATCH] dm: core: Show device sequence instead in dm_dump_tree()) Assuming new revision: v3 ([PATCH] cmd: bind: Use device sequence instead for driver bind/unbind) Analyzing 0 code-review messages
Will use the latest revision: v3
You can pick other revisions using the -vN flag
Checking attestation on all messages, may take a moment...
---
✗ [PATCH] cmd: bind: Use device sequence instead for driver bind/unbind
+ Reviewed-by: Simon Glass <[email protected]> (✗ DKIM/chromium.org)
+ Link: https://lore.kernel.org/r/[email protected]
+ Signed-off-by: Mattijs Korpershoek <[email protected]>
● checkpatch.pl: passed all checks
---
✗ BADSIG: DKIM/hifiphile-com.20230601.gappssmtp.com
---
Total patches: 1
---
Applying: cmd: bind: Use device sequence instead for driver bind/unbind
It applied:
https://patchwork.ozlabs.org/project/uboot/patch/[email protected]/
Instead of:
https://patchwork.ozlabs.org/project/uboot/patch/[email protected]/
I have fixed this up manually, see:
https://source.denx.de/u-boot/custodians/u-boot-dfu/-/commit/73f5b54a589cb6ff97aaca55e7c97d68b43997ee
Quoted 3 lines
--
Mattijs
Re: [PATCH] cmd: bind: Use device sequence instead for driver bind/unbind
In reply to Zixun LIQuoted 8 lines · Zixun LI
Currently uclass index is used for bind/unbind which ignores alias sequence numbering. Use device sequence number instead as it's the number explicitly set in the DT.
Also update documentation to use sequence number.
[...]
Applied to u-boot/next, thanks!
Signature · 1 line
Tom
Quoted 7 lines · Zixun LI
Currently both USB host uclass and USB gadget uclass are using the same name "usb" which break uclass functions like
uclass_get_by_name().Rename the uclass to "
usb_gadget" to fix, also makes bind/unbind by class index (or sequence) working.
Quoted 7 lines
This breaks the capacity of using "usb" as DT alias sequence numbering which needs a fix afterwards.
Signed-off-by Zixun LI <[email protected]>--- drivers/usb/gadget/udc/udc-uclass.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
Hi Zixun,
Thank you for the patch.
Quoted 8 lines · Zixun LI
Currently both USB host uclass and USB gadget uclass are using the same name "usb" which break uclass functions like
uclass_get_by_name().Rename the uclass to "
usb_gadget" to fix, also makes bind/unbind by class index (or sequence) working.This breaks the capacity of using "usb" as DT alias sequence numbering which needs a fix afterwards.
Have you identified boards which use the DT alias that will break with this patch?
Maybe we can detail the required fix in the commit message a bit as well? Or, if you know of a board that uses "usb" as DT alias sequence number, we can submit a fix alongside with this one to document the fix.
Quoted 21 lines
Signed-off-by Zixun LI <[email protected]>--- drivers/usb/gadget/udc/udc-uclass.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)drivers/usb/gadget/udc/udc-uclass.c +1
-2@@ -83,7 +83,7 @@ __weak int dm_usb_gadget_handle_interrupts(struct udevice *dev)#if CONFIG_IS_ENABLED (DM )UCLASS_DRIVER (usb_gadget_generic) = { .id =UCLASS_USB_GADGET_GENERIC , - .name ="usb" , + .name ="usb_gadget" , .flags =DM_UC_FLAG_SEQ_ALIAS , };#endif --2.45.2
Hi Mattijs,
Quoted 8 lines · Mattijs Korpershoek
Have you identified boards which use the DT alias that will break with this patch?
Maybe we can detail the required fix in the commit message a bit as well? Or, if you know of a board that uses "usb" as DT alias sequence number, we can submit a fix alongside with this one to document the fix.
I did a search and it affects mostly i.MX devices and Nuvoton npcm845 where host and device controller use the same alias.
like imx6qdl.dtsi:
aliases {
...
usb0 = &usbotg;
usb1 = &usbh1;
usb2 = &usbh2;
usb3 = &usbh3;
};nuvoton-npcm845-evb.dts:
aliases {
...
usb0 = &udc0;
usb1 = &ehci1;
usb2 = &ehci2;
};But I got lost in EHCI's code, like in ehci_register() when ctrl->init == USB_INIT_DEVICE the function returns without the actual device init ? I'm not sure how their gadget mode works and if it's really broken or not.
Hi Zixun,
Quoted 28 lines · Zixun LI
Hi Mattijs,
Quoted 8 lines · Mattijs Korpershoek
Have you identified boards which use the DT alias that will break with this patch?
Maybe we can detail the required fix in the commit message a bit as well? Or, if you know of a board that uses "usb" as DT alias sequence number, we can submit a fix alongside with this one to document the fix.
I did a search and it affects mostly i.MX devices and Nuvoton npcm845 where host and device controller use the same alias.
like
imx6qdl.dtsi:
aliases {
...usb0 = &usbotg; usb1 = &usbh1; usb2 = &usbh2; usb3 = &usbh3; };
nuvoton-npcm845-evb.dts:
aliases {
...
Quoted 8 lines
usb0 = &udc0; usb1 = &ehci1; usb2 = &ehci2; };But I got lost in EHCI's code, like in
ehci_register()whenctrl->init==USB_INIT_DEVICEthe function returns without the actual device init ? I'm not sure how their gadget mode works and if it's really broken or not.
Thank you for giving some board examples. I am still a bit unclear on the meaning of:
"""
This breaks the capacity of using "usb" as DT alias sequence numbering which needs a fix afterwards.
"""
I have added Jagan, Stefano, Fabio and the NXP team in CC. Does anyone of you have any concerns with this patch ?
If someone could test it, that would be helpful.
Hi Mattijs,
Quoted 9 lines · Mattijs Korpershoek
Thank you for giving some board examples. I am still a bit unclear on the meaning of:
"""
This breaks the capacity of using "usb" as DT alias sequence numbering which needs a fix afterwards.
"""
Quoted 4 lines
I have added Jagan, Stefano, Fabio and the NXP team in CC. Does anyone of you have any concerns with this patch ?
If someone could test it, that would be helpful.
The device sequence number is affected by uclass_find_next_free_seq() in uclass.c, in this function uclass name is used to determine the number.
Since the gadget class's name changed to "usb_gadget" from "usb", alias binding "usb1 = &usbotg;" is not effective anymore. As now it searches for "usb_gadget1 = &usbotg;".
uclass_find_device_by_seq(UCLASS_USB_GADGET_GENERIC,) would fail as sequence number is changed.
I made a tentative fix in
https://lore.kernel.org/u-boot/[email protected]/
But Simon doesn't want to modify uclass_driver structure only for gadget class.
Regards,
Zixun
Mattijs Korpershoek
Reviewed-byHi Zixun,
Quoted 21 lines · Zixun LI
Hi Mattijs,
Quoted 9 lines · Mattijs Korpershoek
Thank you for giving some board examples. I am still a bit unclear on the meaning of:
"""
This breaks the capacity of using "usb" as DT alias sequence numbering which needs a fix afterwards.
"""Quoted 4 lines
I have added Jagan, Stefano, Fabio and the NXP team in CC. Does anyone of you have any concerns with this patch ?
If someone could test it, that would be helpful.
The device sequence number is affected by
uclass_find_next_free_seq()inuclass.c, in this function uclass name is used to determine the number.
Quoted 6 lines
Since the gadget class's name changed to "
usb_gadget" from "usb", alias binding "usb1 = &usbotg;" is not effective anymore. As now it searches for "usb_gadget1= &usbotg;".
uclass_find_device_by_seq(UCLASS_USB_GADGET_GENERIC,)would fail as sequence number is changed.
Thank you for the explanation.
Looking at the code, I don't see anyone calling:
- uclass_find_device_by_seq(UCLASS_USB_GADGET
And one occurrence of:
- uclass_get_device_by_seq(UCLASS_USB_GADGET_GENERIC, index, &dev);
Quoted 8 lines
I made a tentative fix in
https://lore.kernel.org/u-boot/[email protected]/
But Simon doesn't want to modifyuclass_driverstructure only for gadget class.Regards,
Zixun