musb: mediatek: rename driver

Commit Message

Frank Wunderlich April 30, 2021, 12:43 p.m. UTC

From: Frank Wunderlich <[e-mail]>

currently unspecific mediatek.ko is built,
change this by adding subsystem

Signed-off-by: Frank Wunderlich <[e-mail]>
---
 drivers/usb/musb/Makefile                   | 2 +-
 drivers/usb/musb/{mediatek.c => musb_mtk.c} | 0
 2 files changed, 1 insertion(+), 1 deletion(-)
 rename drivers/usb/musb/{mediatek.c => musb_mtk.c} (100%)

Comments

Frank Wunderlich April 30, 2021, 1:30 p.m. UTC | #1

Hi Grek

the problem is that the name mediatek.ko does not point to musb-subsystem. I had discussed with Min Gao (author of this driver) some time ago about this as the name may conflict with other modules (don't remember which was that).
We have searched issue using the driver on my board (not yet resolved).

if the module is loaded it shows (based on name) only "mediatek" and user does not know that is the mediatek musb driver, not very good in my eyes as mediatek is a vendor designing many different hardware and so drivers. Imho the module-name should at least give a clue to the subsystem to avoid confusion/conflicts

Now the discussion comes up again here for a new driver:
https://patchwork.kernel.org/project/linux-mediatek/cover/[e-mail]/#24148777

so i decided to rebase and post my patch created in past to clean this up.

and yes this can result in user-space issues depending on the name...because of this i have not added stable-tag ;)

regards Frank


> Gesendet: Freitag, 30. April 2021 um 14:54 Uhr
> Von: "Greg Kroah-Hartman" <[e-mail]>

> On Fri, Apr 30, 2021 at 02:43:17PM +0200, Frank Wunderlich wrote:
> > From: Frank Wunderlich <[e-mail]>
> >
> > currently unspecific mediatek.ko is built,
> > change this by adding subsystem
>
> I am sorry, I do not understand this changelog text.  What are you doing
> here and why?
>
> > Signed-off-by: Frank Wunderlich <[e-mail]>
> > ---
> >  drivers/usb/musb/Makefile                   | 2 +-
> >  drivers/usb/musb/{mediatek.c => musb_mtk.c} | 0
>
> Why rename this file, will that not break existing setups that expect
> the module to be named this?
>
> thanks,
>
> greg k-h

Greg KH April 30, 2021, 1:57 p.m. UTC | #2

Please do not top-post :)

On Fri, Apr 30, 2021 at 03:30:39PM +0200, Frank Wunderlich wrote:
> Hi Grek
> 
> the problem is that the name mediatek.ko does not point to musb-subsystem. I had discussed with Min Gao (author of this driver) some time ago about this as the name may conflict with other modules (don't remember which was that).
> We have searched issue using the driver on my board (not yet resolved).
> 
> if the module is loaded it shows (based on name) only "mediatek" and user does not know that is the mediatek musb driver, not very good in my eyes as mediatek is a vendor designing many different hardware and so drivers. Imho the module-name should at least give a clue to the subsystem to avoid confusion/conflicts
> 
> Now the discussion comes up again here for a new driver:
> https://patchwork.kernel.org/project/linux-mediatek/cover/[e-mail]/#24148777
> 
> so i decided to rebase and post my patch created in past to clean this up.

All of this information needs to go into the changelog text, what you
wrote there does not explain all of the above.

> and yes this can result in user-space issues depending on the name...because of this i have not added stable-tag ;)

Why does stable matter?  If this could result in a userspace breakage,
that's not ok for any kernel change.

Why not just have any new driver not use this name, as it is, it's not
hurting anything.  Until you can guarantee that renaming this is not
going to break anything, I can not take it.

thanks,

greg k-h

Frank Wunderlich April 30, 2021, 2:38 p.m. UTC | #3

Am 30. April 2021 15:57:17 MESZ schrieb Greg Kroah-Hartman <[e-mail]>:

>Why not just have any new driver not use this name, as it is, it's not
>hurting anything.

I just want to make it clean for users/admins for what kind of device this module is for (vendor.ko is imho not good at least for vendors manufacturing such different devices).
And avoid such conflicts like the phy (which had reintroduced the unclear module name).
The new module will be renamed in next version..there is only discussion about the new name...and so i tried to fix the "other end".

>Until you can guarantee that renaming this is not
>going to break anything, I can not take it.

If a user/administrator installs a new major kernel version he needs to check if all is still working. Then the new module can be loaded...i don't see a real breakage or much work for "fix" this,but it's ok if you do not take it up.

This driver is afair available since 5.4 so i guess not much spreading in large productive environments mostly still using 4.14 or 4.4

I can repost with better description,but this is not needed if change itself will be rejected. So maybe anyone else can comment too.

regards Frank

Patch

diff --git a/drivers/usb/musb/Makefile b/drivers/usb/musb/Makefile
index 932247360a9f..82928d4fd42a 100644
--- a/drivers/usb/musb/Makefile
+++ b/drivers/usb/musb/Makefile
@@ -24,7 +24,7 @@  obj-$(CONFIG_USB_MUSB_DA8XX)			+= da8xx.o
 obj-$(CONFIG_USB_MUSB_UX500)			+= ux500.o
 obj-$(CONFIG_USB_MUSB_JZ4740)			+= jz4740.o
 obj-$(CONFIG_USB_MUSB_SUNXI)			+= sunxi.o
-obj-$(CONFIG_USB_MUSB_MEDIATEK)      		+= mediatek.o
+obj-$(CONFIG_USB_MUSB_MEDIATEK)      		+= musb_mtk.o
 
 # the kconfig must guarantee that only one of the
 # possible I/O schemes will be enabled at a time ...
diff --git a/drivers/usb/musb/mediatek.c b/drivers/usb/musb/musb_mtk.c
similarity index 100%
rename from drivers/usb/musb/mediatek.c
rename to drivers/usb/musb/musb_mtk.c