| Message ID | 20211004141839.49079-1-andriy.shevchenko@linux.intel.com (mailing list archive) |
|---|---|
| State | Accepted |
| Commit | 268bbde716e3a79a747a0f4ebbeb9f63d861737d |
| Headers | show |
| Series | [v1,1/1] usb: dwc3: gadget: Revert "set gadgets parent to the right controller" | expand |
Hi, Op 04-10-2021 om 16:18 schreef Andy Shevchenko: > The commit c6e23b89a95d ("usb: dwc3: gadget: set gadgets parent to the right > controller") changed the device for the UDC and broke the user space scripts > that instantiate the USB gadget(s) via ConfigFS. I confirm this regression on Intel Edison since at least 5.15-rc2 while in 5.14.0 it was working fine. This patch resolves the issue as tested on 5.15-rc4. Tested-by: Ferry Toth<fntoth@gmail.com> > Revert it for now until the better solution will be proposed. > > Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com> > --- > drivers/usb/dwc3/gadget.c | 2 +- > 1 file changed, 1 insertion(+), 1 deletion(-) > > diff --git a/drivers/usb/dwc3/gadget.c b/drivers/usb/dwc3/gadget.c > index 8e66a70adae6..13664609ed3c 100644 > --- a/drivers/usb/dwc3/gadget.c > +++ b/drivers/usb/dwc3/gadget.c > @@ -4253,7 +4253,7 @@ int dwc3_gadget_init(struct dwc3 *dwc) > } > > > - usb_initialize_gadget(dwc->sysdev, dwc->gadget, dwc_gadget_release); > + usb_initialize_gadget(dwc->dev, dwc->gadget, dwc_gadget_release); > dev = &dwc->gadget->dev; > dev->platform_data = dwc; > dwc->gadget->ops = &dwc3_gadget_ops;
On Mon, Oct 04, 2021 at 10:35:57PM +0200, Ferry Toth wrote: >Hi, > >Op 04-10-2021 om 16:18 schreef Andy Shevchenko: >>The commit c6e23b89a95d ("usb: dwc3: gadget: set gadgets parent to the right >>controller") changed the device for the UDC and broke the user space scripts >>that instantiate the USB gadget(s) via ConfigFS. > >I confirm this regression on Intel Edison since at least 5.15-rc2 >while in 5.14.0 it was working fine. > >This patch resolves the issue as tested on 5.15-rc4. > >Tested-by: Ferry Toth<fntoth@gmail.com> NACK! Why should we resolv an issue by reverting it to solve not working userspace. We already have this patch as a solution for solving a deeper Problem, regarding the allocator addressing the right device. >>Revert it for now until the better solution will be proposed. So, I think fixing the userspace would be the right fix, not changing the kernel. Otherwise we should find a proper solution. >> >>Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com> >>--- >> drivers/usb/dwc3/gadget.c | 2 +- >> 1 file changed, 1 insertion(+), 1 deletion(-) >> >>diff --git a/drivers/usb/dwc3/gadget.c b/drivers/usb/dwc3/gadget.c >>index 8e66a70adae6..13664609ed3c 100644 >>--- a/drivers/usb/dwc3/gadget.c >>+++ b/drivers/usb/dwc3/gadget.c >>@@ -4253,7 +4253,7 @@ int dwc3_gadget_init(struct dwc3 *dwc) >> } >>- usb_initialize_gadget(dwc->sysdev, dwc->gadget, dwc_gadget_release); >>+ usb_initialize_gadget(dwc->dev, dwc->gadget, dwc_gadget_release); >> dev = &dwc->gadget->dev; >> dev->platform_data = dwc; >> dwc->gadget->ops = &dwc3_gadget_ops; >
On Tue, Oct 5, 2021 at 11:52 AM Michael Grzeschik <mgr@pengutronix.de> wrote: > On Mon, Oct 04, 2021 at 10:35:57PM +0200, Ferry Toth wrote: > >Op 04-10-2021 om 16:18 schreef Andy Shevchenko: > >>The commit c6e23b89a95d ("usb: dwc3: gadget: set gadgets parent to the right > >>controller") changed the device for the UDC and broke the user space scripts > >>that instantiate the USB gadget(s) via ConfigFS. > > > >I confirm this regression on Intel Edison since at least 5.15-rc2 > >while in 5.14.0 it was working fine. > > > >This patch resolves the issue as tested on 5.15-rc4. > > > >Tested-by: Ferry Toth<fntoth@gmail.com> > > NACK! Why should we resolv an issue by reverting it to solve not working > userspace. Huh?! It is a) used to work; b) stopped working after your commit. To me it's a clear regression. Whatever deeper problem is there, I really don't care. The change broke my _user space_ case! > We already have this patch as a solution for solving a deeper > Problem, regarding the allocator addressing the right device. Then rework it. You have still time. Your case wasn't working and one more release of not working is not an issue here. > >>Revert it for now until the better solution will be proposed. > > So, I think fixing the userspace would be the right fix, Huh?! https://www.kernel.org/doc/html/latest/process/4.Coding.html#regressions > not changing > the kernel. Otherwise we should find a proper solution. So, please do. v5.15 should still work on our devices with distribution that uses ConfigFS, no?
Michael Grzeschik <mgr@pengutronix.de> writes: > [[PGP Signed Part:Undecided]] > On Mon, Oct 04, 2021 at 10:35:57PM +0200, Ferry Toth wrote: >>Hi, >> >>Op 04-10-2021 om 16:18 schreef Andy Shevchenko: >>>The commit c6e23b89a95d ("usb: dwc3: gadget: set gadgets parent to the right >>>controller") changed the device for the UDC and broke the user space scripts >>>that instantiate the USB gadget(s) via ConfigFS. >> >> I confirm this regression on Intel Edison since at least 5.15-rc2 >> while in 5.14.0 it was working fine. >> >>This patch resolves the issue as tested on 5.15-rc4. >> >>Tested-by: Ferry Toth<fntoth@gmail.com> > > NACK! Why should we resolv an issue by reverting it to solve not working > userspace. We already have this patch as a solution for solving a deeper heh, there is only one rule in this community: thou shalt not break userspace :-)
On Tue, Oct 05, 2021 at 10:51:00AM +0200, Michael Grzeschik wrote: > On Mon, Oct 04, 2021 at 10:35:57PM +0200, Ferry Toth wrote: > > Hi, > > > > Op 04-10-2021 om 16:18 schreef Andy Shevchenko: > > > The commit c6e23b89a95d ("usb: dwc3: gadget: set gadgets parent to the right > > > controller") changed the device for the UDC and broke the user space scripts > > > that instantiate the USB gadget(s) via ConfigFS. > > > > I confirm this regression on Intel Edison since at least 5.15-rc2 while > > in 5.14.0 it was working fine. > > > > This patch resolves the issue as tested on 5.15-rc4. > > > > Tested-by: Ferry Toth<fntoth@gmail.com> > > NACK! Why should we resolv an issue by reverting it to solve not working > userspace. We already have this patch as a solution for solving a deeper > Problem, regarding the allocator addressing the right device. > > > > Revert it for now until the better solution will be proposed. > > So, I think fixing the userspace would be the right fix, not changing > the kernel. Otherwise we should find a proper solution. We only really have one rule in Linux kernel development: If a kernel change breaks userspace, the kernel change needs to be reverted. Go fix up the userspace tools first, ensure everyone has updated, and then we can consider taking the change back into the kernel tree. thanks, greg k-h
diff --git a/drivers/usb/dwc3/gadget.c b/drivers/usb/dwc3/gadget.c index 8e66a70adae6..13664609ed3c 100644 --- a/drivers/usb/dwc3/gadget.c +++ b/drivers/usb/dwc3/gadget.c @@ -4253,7 +4253,7 @@ int dwc3_gadget_init(struct dwc3 *dwc) } - usb_initialize_gadget(dwc->sysdev, dwc->gadget, dwc_gadget_release); + usb_initialize_gadget(dwc->dev, dwc->gadget, dwc_gadget_release); dev = &dwc->gadget->dev; dev->platform_data = dwc; dwc->gadget->ops = &dwc3_gadget_ops;
The commit c6e23b89a95d ("usb: dwc3: gadget: set gadgets parent to the right controller") changed the device for the UDC and broke the user space scripts that instantiate the USB gadget(s) via ConfigFS. Revert it for now until the better solution will be proposed. Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com> --- drivers/usb/dwc3/gadget.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)