On Sun, Feb 06, 2022 at 12:51:00PM +0000, Sa, Nuno wrote: > > From: Andy Shevchenko <andriy.shevchenko@xxxxxxxxxxxxxxx> > > Sent: Saturday, February 5, 2022 7:24 PM > > On Sat, Feb 05, 2022 at 05:14:54PM +0000, Jonathan Cameron wrote: > > > On Thu, 3 Feb 2022 13:45:06 +0200 > > > Andy Shevchenko <andriy.shevchenko@xxxxxxxxxxxxxxx> wrote: ... > > > > +#include <asm/byteorder.h> > > > > +#include <asm/unaligned.h> > > > > > This may well be a valid addition but it's not called out in the patch > > > description. > > > > This is a side effect of the change. Below I will try to explain, tell me if > > that is what you want to be added to the commit message (feel free to > > correct > > my English). > > > > The conversion slightly changes the logic behind property reading for > > the > > configuration values. Original code allocates just as much memory as > > needed. > > Then for each separate 32- or 64-bit value it reads it from the property > > and converts to a raw one which will be fed to the sensor. In the new > > code > > we allocated the amount of memory needed to retrieve all values at > > once from > > the property and then convert them as required. > > Yeah, I understood what you are doing... We allocate some extra bytes so > that we can simplify a lot the logic to get, convert and put back the properties > in the table... And since we are only writing the __real__ table size into the device > I'm fairly positive this works. However, I do agree with Jonathan and I would > be more confident to give this a test before being applied... > > That said, I'm off till the end of the month (just doing minor stuff like replying > to emails :)), so if we can hold this one, I will test it first thing I'm back in the > office. Will wait for your testing. Meanwhile I address all other comments and send v3 for further review. -- With Best Regards, Andy Shevchenko
