Re: Possible big endian bug in latest stable kernel

[Greg KH <gregkh@linuxfoundation.org>]

On Thu, Jan 27, 2022 at 05:33:27PM +0000, Rory Bolt wrote:
> 
> Hello Greg,
> 
> I used bisect to find the offending commit:
> 
> To my surprise, it appears to be only peripherally related to USB; it is in fact the change to drivers/regulators/core.c
> 
> I cannot explain how this only affects big endian operation, not little endian. I can only theorize that since the u-boot firmware is in little endian mode, it has already set up the regulators... however when I transition to big endian mode, I essentially re-init everything.
> 
> aarch64_be-test /usr/src/linux-5.15.14/drivers/regulator # diff core.c.orig core.c
> 1531c1531,1532
> <               if (rdev->supply_name && !rdev->supply)
> ---
> >               if (rdev->supply_name && !rdev->supply &&
> >                   !rdev->constraints->always_on)
> 
> I am uncertain if this patch is generally correct or not, hence my reluctance to post. It seems that rdev-supply is ALWAYS null and rdev->constraints->always_on is ALWAYS true for the RockPro64 board. My concern is that this may affect other platforms.
> 
> Should I post this to the general kernel mailing list? Is there another more appropriate list?

You point at a merge commit, but not the actual commit that causes the
problem within that merge commit.  I would cc: the developers involved
in that tree and point them at this git bisect run and work with them to
resolve this issue.

thanks,

greg k-h