On Sun, 27 Oct 2019 at 14:39, Greg KH greg@kroah.com wrote:
On Sat, Oct 26, 2019 at 05:46:03PM +0200, Ard Biesheuvel wrote:
On Sat, 26 Oct 2019 at 17:40, Sasha Levin sashal@kernel.org wrote:
On Sat, Oct 26, 2019 at 10:01:21AM +0200, Greg KH wrote:
On Fri, Oct 25, 2019 at 05:39:44PM +0200, Ard Biesheuvel wrote:
On Fri, 25 Oct 2019 at 17:28, Ard Biesheuvel ard.biesheuvel@linaro.org wrote:
On Fri, 25 Oct 2019 at 17:25, Sasha Levin sashal@kernel.org wrote: > > On Thu, Oct 24, 2019 at 04:37:12PM +0200, Ard Biesheuvel wrote: > >On Thu, 24 Oct 2019 at 16:34, Alexandru Elisei alexandru.elisei@arm.com wrote:
...
> >> > >> This breaks when building, because __hardenbp_enab is declared in the next patch: > >> > >> $ make -j32 defconfig && make -j32 > >> > >> [..] > >> arch/arm64/kernel/cpu_errata.c: In function ‘check_branch_predictor’: > >> arch/arm64/kernel/cpu_errata.c:492:3: error: ‘__hardenbp_enab’ undeclared (first > >> use in this function) > >> __hardenbp_enab = false; > >> ^~~~~~~~~~~~~~~ > >> arch/arm64/kernel/cpu_errata.c:492:3: note: each undeclared identifier is reported > >> only once for each function it appears in > >> make[1]: *** [scripts/Makefile.build:326: arch/arm64/kernel/cpu_errata.o] Error 1 > >> make[1]: *** Waiting for unfinished jobs.... > >> > > > >Indeed, but as discussed, this matches the state of both mainline and > >v4.19, which carry these patches in the same [wrong] order as well. > > > >Greg should confirm, but as I understand it, it is preferred to be > >bug-compatible with mainline rather than fixing problems when spotting > >them while doing the backport. > > Is it just patch ordering? If so I'd rather fix it, there's no reason to > carry this issue into the stable trees. > > We reserve "bug compatibility" for functional issues that are not yet > fixed upstream, it doesn't seem to be the case here. >
The patches don't apply cleanly in the opposite order.
What we could do is squash the two patches together. That way, we avoid the breakage without having to modify the patches in order to be able to apply them.
No, don't do that. Just take all of the needed commits.
Right, just make the patches apply in reverse, this shouldn't be more than moving some code from the 2nd patch back to the 1st one, right?
We usually don't do this in stable backports, but there are three good reasons to do it here:
- It'll be nice to maintain bisectability.
- The end result should be exactly the same, so there's no room for
error here. 3. It's a backport for an older kernel to begin with, so there are changes from upstream already.
I really don't see the point of doing this for v4.14 while v4.19 and mainline have the two patches in the opposite order.
Would you like me to resend the entire 48 piece series for this?
No need, I've queued the whole thing up now, thanks.
Thanks Greg