Hi Greg, Sasha, and Michael,
Commit d79976918852 ("powerpc/64: Add UADDR64 relocation support") fixes a boot failure with CONFIG_RELOCATABLE=y kernels linked with recent versions of ld.lld [1]. Additionally, it resolves a separate boot failure that Paul Menzel reported [2] with ld.lld 13.0.0. Is this a reasonable backport for 5.17 and 5.15? It applies cleanly, resolves both problems, and does not appear to cause any other issues in my testing for both trees but I was curious what Michael's opinion was, as I am far from a PowerPC expert.
This change does apply cleanly to 5.10 (I did not try earlier branches) but there are other changes needed for ld.lld to link CONFIG_RELOCATABLE kernels in that branch so to avoid any regressions, I think it is safe to just focus on 5.15 and 5.17.
Paul, it would not hurt to confirm the results of my testing with your setup, just to make sure I did not miss anything :)
[1]: https://github.com/ClangBuiltLinux/linux/issues/1581 [2]: https://lore.kernel.org/Yg2h2Q2vXFkkLGTh@dev-arch.archlinux-ax161/
Cheers, Nathan
Nathan Chancellor nathan@kernel.org writes:
Hi Greg, Sasha, and Michael,
Commit d79976918852 ("powerpc/64: Add UADDR64 relocation support") fixes a boot failure with CONFIG_RELOCATABLE=y kernels linked with recent versions of ld.lld [1]. Additionally, it resolves a separate boot failure that Paul Menzel reported [2] with ld.lld 13.0.0. Is this a reasonable backport for 5.17 and 5.15? It applies cleanly, resolves both problems, and does not appear to cause any other issues in my testing for both trees but I was curious what Michael's opinion was, as I am far from a PowerPC expert.
This change does apply cleanly to 5.10 (I did not try earlier branches) but there are other changes needed for ld.lld to link CONFIG_RELOCATABLE kernels in that branch so to avoid any regressions, I think it is safe to just focus on 5.15 and 5.17.
I considered tagging it for stable, but I wanted it to get a bit of testing first, it's a reasonably big patch.
I think we're reasonably confident it doesn't introduce any new bugs, but more testing time is always good.
So I guess I'd be inclined to wait another week or so before requesting a stable backport?
cheers
On Thu, Apr 21, 2022 at 05:46:52PM +1000, Michael Ellerman wrote:
Nathan Chancellor nathan@kernel.org writes:
Hi Greg, Sasha, and Michael,
Commit d79976918852 ("powerpc/64: Add UADDR64 relocation support") fixes a boot failure with CONFIG_RELOCATABLE=y kernels linked with recent versions of ld.lld [1]. Additionally, it resolves a separate boot failure that Paul Menzel reported [2] with ld.lld 13.0.0. Is this a reasonable backport for 5.17 and 5.15? It applies cleanly, resolves both problems, and does not appear to cause any other issues in my testing for both trees but I was curious what Michael's opinion was, as I am far from a PowerPC expert.
This change does apply cleanly to 5.10 (I did not try earlier branches) but there are other changes needed for ld.lld to link CONFIG_RELOCATABLE kernels in that branch so to avoid any regressions, I think it is safe to just focus on 5.15 and 5.17.
I considered tagging it for stable, but I wanted it to get a bit of testing first, it's a reasonably big patch.
I think we're reasonably confident it doesn't introduce any new bugs, but more testing time is always good.
So I guess I'd be inclined to wait another week or so before requesting a stable backport?
Sure, thanks for the response! I'll ping this thread on Monday, May 2nd, so that we have two more RC releases to try and flush out any lingering issues. If you do receive any reports of regressions, please let me know.
Cheers, Nathan
Nathan Chancellor nathan@kernel.org writes:
On Thu, Apr 21, 2022 at 05:46:52PM +1000, Michael Ellerman wrote:
Nathan Chancellor nathan@kernel.org writes:
Hi Greg, Sasha, and Michael,
Commit d79976918852 ("powerpc/64: Add UADDR64 relocation support") fixes a boot failure with CONFIG_RELOCATABLE=y kernels linked with recent versions of ld.lld [1]. Additionally, it resolves a separate boot failure that Paul Menzel reported [2] with ld.lld 13.0.0. Is this a reasonable backport for 5.17 and 5.15? It applies cleanly, resolves both problems, and does not appear to cause any other issues in my testing for both trees but I was curious what Michael's opinion was, as I am far from a PowerPC expert.
This change does apply cleanly to 5.10 (I did not try earlier branches) but there are other changes needed for ld.lld to link CONFIG_RELOCATABLE kernels in that branch so to avoid any regressions, I think it is safe to just focus on 5.15 and 5.17.
I considered tagging it for stable, but I wanted it to get a bit of testing first, it's a reasonably big patch.
I think we're reasonably confident it doesn't introduce any new bugs, but more testing time is always good.
So I guess I'd be inclined to wait another week or so before requesting a stable backport?
Sure, thanks for the response! I'll ping this thread on Monday, May 2nd, so that we have two more RC releases to try and flush out any lingering issues. If you do receive any reports of regressions, please let me know.
Sounds good, thanks.
cheers
On Thu, Apr 21, 2022 at 08:13:13AM -0700, Nathan Chancellor wrote:
On Thu, Apr 21, 2022 at 05:46:52PM +1000, Michael Ellerman wrote:
Nathan Chancellor nathan@kernel.org writes:
Hi Greg, Sasha, and Michael,
Commit d79976918852 ("powerpc/64: Add UADDR64 relocation support") fixes a boot failure with CONFIG_RELOCATABLE=y kernels linked with recent versions of ld.lld [1]. Additionally, it resolves a separate boot failure that Paul Menzel reported [2] with ld.lld 13.0.0. Is this a reasonable backport for 5.17 and 5.15? It applies cleanly, resolves both problems, and does not appear to cause any other issues in my testing for both trees but I was curious what Michael's opinion was, as I am far from a PowerPC expert.
This change does apply cleanly to 5.10 (I did not try earlier branches) but there are other changes needed for ld.lld to link CONFIG_RELOCATABLE kernels in that branch so to avoid any regressions, I think it is safe to just focus on 5.15 and 5.17.
I considered tagging it for stable, but I wanted it to get a bit of testing first, it's a reasonably big patch.
I think we're reasonably confident it doesn't introduce any new bugs, but more testing time is always good.
So I guess I'd be inclined to wait another week or so before requesting a stable backport?
Sure, thanks for the response! I'll ping this thread on Monday, May 2nd, so that we have two more RC releases to try and flush out any lingering issues. If you do receive any reports of regressions, please let me know.
I decided to wait an extra day just to give people the opportunity to install -rc5 and run it through their tests. I have not heard of any reports yet, are there any further objections?
Cheers, Nathan
Nathan Chancellor nathan@kernel.org writes:
On Thu, Apr 21, 2022 at 08:13:13AM -0700, Nathan Chancellor wrote:
On Thu, Apr 21, 2022 at 05:46:52PM +1000, Michael Ellerman wrote:
Nathan Chancellor nathan@kernel.org writes:
Hi Greg, Sasha, and Michael,
Commit d79976918852 ("powerpc/64: Add UADDR64 relocation support") fixes a boot failure with CONFIG_RELOCATABLE=y kernels linked with recent versions of ld.lld [1]. Additionally, it resolves a separate boot failure that Paul Menzel reported [2] with ld.lld 13.0.0. Is this a reasonable backport for 5.17 and 5.15? It applies cleanly, resolves both problems, and does not appear to cause any other issues in my testing for both trees but I was curious what Michael's opinion was, as I am far from a PowerPC expert.
This change does apply cleanly to 5.10 (I did not try earlier branches) but there are other changes needed for ld.lld to link CONFIG_RELOCATABLE kernels in that branch so to avoid any regressions, I think it is safe to just focus on 5.15 and 5.17.
I considered tagging it for stable, but I wanted it to get a bit of testing first, it's a reasonably big patch.
I think we're reasonably confident it doesn't introduce any new bugs, but more testing time is always good.
So I guess I'd be inclined to wait another week or so before requesting a stable backport?
Sure, thanks for the response! I'll ping this thread on Monday, May 2nd, so that we have two more RC releases to try and flush out any lingering issues. If you do receive any reports of regressions, please let me know.
I decided to wait an extra day just to give people the opportunity to install -rc5 and run it through their tests. I have not heard of any reports yet, are there any further objections?
No objection.
cheers
On Wed, May 04, 2022 at 01:19:32PM +1000, Michael Ellerman wrote:
Nathan Chancellor nathan@kernel.org writes:
On Thu, Apr 21, 2022 at 08:13:13AM -0700, Nathan Chancellor wrote:
On Thu, Apr 21, 2022 at 05:46:52PM +1000, Michael Ellerman wrote:
Nathan Chancellor nathan@kernel.org writes:
Hi Greg, Sasha, and Michael,
Commit d79976918852 ("powerpc/64: Add UADDR64 relocation support") fixes a boot failure with CONFIG_RELOCATABLE=y kernels linked with recent versions of ld.lld [1]. Additionally, it resolves a separate boot failure that Paul Menzel reported [2] with ld.lld 13.0.0. Is this a reasonable backport for 5.17 and 5.15? It applies cleanly, resolves both problems, and does not appear to cause any other issues in my testing for both trees but I was curious what Michael's opinion was, as I am far from a PowerPC expert.
This change does apply cleanly to 5.10 (I did not try earlier branches) but there are other changes needed for ld.lld to link CONFIG_RELOCATABLE kernels in that branch so to avoid any regressions, I think it is safe to just focus on 5.15 and 5.17.
I considered tagging it for stable, but I wanted it to get a bit of testing first, it's a reasonably big patch.
I think we're reasonably confident it doesn't introduce any new bugs, but more testing time is always good.
So I guess I'd be inclined to wait another week or so before requesting a stable backport?
Sure, thanks for the response! I'll ping this thread on Monday, May 2nd, so that we have two more RC releases to try and flush out any lingering issues. If you do receive any reports of regressions, please let me know.
I decided to wait an extra day just to give people the opportunity to install -rc5 and run it through their tests. I have not heard of any reports yet, are there any further objections?
No objection.
Now queued up!
linux-stable-mirror@lists.linaro.org
-
Greg Kroah-Hartman -
Michael Ellerman -
Nathan Chancellor