On Thu, Aug 01, 2019 at 08:34:45AM +0100, Will Deacon wrote:
On Thu, Aug 01, 2019 at 12:35:41PM +0530, Viresh Kumar wrote:
On 01-08-19, 08:57, Greg KH wrote:
On Thu, Aug 01, 2019 at 12:05:44PM +0530, Viresh Kumar wrote:
On 01-08-19, 07:30, Julien Thierry wrote:
I must admit I am not familiar with backport/stable process enough. But personally I think the your suggestion seems more sensible than backporting 4 patches.
Or you can maybe ignore patch 25 and say in patch 24 that among the changes made for the 4.4 codebase, the call arm64_apply_bp_hardening() was moved from post_ttbr_update_workaround as it doesn't exist and placed in check_and_switch_context() as it is its final destination.
Done that and dropped the other two patches.
However, I really don't know what's the best way to proceed according to existing practices. So input from someone else would be welcome.
Lets see if someone comes up and ask me to do something else :)
Keeping the same patches that upstream has is almost always the better thing to do in the long-run.
That would require two additional patches to be backported, 22 and 23 from this series. From your suggestion it seems that keeping them is better here ?
Yes. Backporting individual patches as they appear upstream is definitely the preferred method for -stable. It makes the relationship to mainline crystal clear, as well as any dependencies between patches that have been backported. Everytime we tweak something unecessarily in a stable backport, it just creates the potential for confusion and additional conflicts in future backports, so it's best to follow the shape of upstream as closely as possible, even if it results in additional patches.
So I wouldn't worry about total number of patches. I'd worry more about things like conflicts, deviation from mainline and overall testing coverage.
That is exactly correct, thanks for saying it better than I could :)
greg k-h