On Fri, 15 Apr 2022, Greg Kroah-Hartman wrote:
> > Remove redundant code then and factor out OxSemi Tornado device
> > detection. Also correct the baud base like with commit 6cbe45d8ac93
> > ("serial: 8250: Correct the clock for OxSemi PCIe devices") for the
> > value of 3906250 rather than 4000000, obtained by dividing the 62.5MHz
> > clock input by the default oversampling rate of 16. Finally move the
> > EndRun vendor:device ID to <linux/pci_ids.h>.
>
> That's a lot of different things happening all the same commit. Please
> break this out into one-patch-per-logical-change as is required.
The baud base fix is completely swallowed by the next change for EndRun
devices, but I guess someone may want to backport it on its own, however
unlikely.
I have posted v4 then with this change split off (and the other removed)
as per your request. I have also reconsidered the changes made in 2/2 and
split it into three, so that drivers/tty/serial/8250/8250_port.c updates
are separate and self-contained.
In the course of the respin, I have realised exporting the ICR access
helpers caused a code generation regression, so I have removed the inline
function specifier so as to let the compiler choose whether to inline the
functions or not. I have also realised that the change to the console
restorer is actually a fix for a preexisting bug in handling of the AFE
bit, so I have annotated the change accordingly.
Thank you for your review.
Maciej
