On Thu, Aug 08, 2019 at 05:36:00PM +0530, Viresh Kumar wrote:
On 06-08-19, 13:18, Mark Rutland wrote:
Upstream and in v4.9, the meltdown patches came before the spectre patches, and doing this in the opposite order causes context problems like the above.
Given that, I think it would be less surprising to do the meltdown backport first, though I apprecaite that's more work to get these patches in. :/
I attempted meltdown backport in the last two days and the amount of extra patches to be backported is enormous. And I am not sure if everything is alright as well now, and things will greatly rely on reviews from you for it.
For this series, what about just backporting for now to account for CSV3 ? And attempting meltdown backport separately later ?
179a56f6f9fb arm64: Take into account ID_AA64PFR0_EL1.CSV3
I don't think that buys us anything; that's still going to cause some context problems (e.g. for commit 179a56f6f9fb itself), and still means that the v4.4 backport differs from all the others.
If it's really not feasible to do the meltdown patches first, then I reluctantly agree that we should just do the spectre bits alone if there aren't major changes that have to be made to entry.S and friends as a result.
Could you send a v3 (of just the spectre bits) with the changes requested so far?
Thanks, Mark.