| 1 |
On 01/23/2018 11:06 PM, Adam Carter wrote: |
| 2 |
> |
| 3 |
> FYI, this is from 4.14.15's ChangeLog, doesn't appear to be in 4.9.78 |
| 4 |
> though. |
| 5 |
> |
| 6 |
> commit f41b2d7ee791e845147a3e3cafb25493dca6870a |
| 7 |
> Author: Josh Poimboeuf <jpoimboe@××××××.com <mailto:jpoimboe@××××××.com>> |
| 8 |
> Date: Mon Jan 15 08:17:07 2018 -0600 |
| 9 |
> |
| 10 |
> objtool: Fix seg fault with gold linker |
| 11 |
> |
| 12 |
> commit 2a0098d70640dda192a79966c14d449e7a34d675 upstream. |
| 13 |
> |
| 14 |
> Objtool segfaults when the gold linker is used with |
| 15 |
> CONFIG_MODVERSIONS=y and CONFIG_UNWINDER_ORC=y. |
| 16 |
> |
| 17 |
> With CONFIG_MODVERSIONS=y, the .o file gets passed to the linker before |
| 18 |
> being passed to objtool. The gold linker seems to strip unused ELF |
| 19 |
> symbols by default, which confuses objtool and causes the seg fault when |
| 20 |
> it's trying to generate ORC metadata. |
| 21 |
> |
| 22 |
> Objtool should really be running immediately after GCC anyway, without a |
| 23 |
> linker call in between. Change the makefile ordering so that objtool is |
| 24 |
> called before the linker. |
| 25 |
. |
| 26 |
FYI : |
| 27 |
Found the culprit ... ( very late last night ) |
| 28 |
. |
| 29 |
sys-kernel/gentoo-sources has a USE Flag of 'experimental' |
| 30 |
The experimental patch enables additional choices for CPU. |
| 31 |
The last kernel version with no problem was 4.9.76-r1 |
| 32 |
The choice that worked before was : AMD Piledriver |
| 33 |
That induces the error now. |
| 34 |
. |
| 35 |
When I compiled with CPU set to : |
| 36 |
K-8 --> no problem |
| 37 |
Generic x86_64 --> no problem |
| 38 |
. |
| 39 |
That was the only change made in 'menuconfig' before every test compile. |
| 40 |
. |
| 41 |
Now I know why all the warnings from the 'objtool', during compile, also |
| 42 |
disappeared. |
| 43 |
( With CPU set to either : K-8, Generic x86_64 ) |
| 44 |
> Objtool should really be running immediately after GCC anyway, without a |
| 45 |
> linker call in between. Change the makefile ordering so that objtool is |
| 46 |
> called before the linker. |
| 47 |
Thinking the 'experimental patch' may be needing some attention. |
| 48 |
. |
| 49 |
So ... now I am looking at filling a bug report ( if appropriate ) on |
| 50 |
Gentoo's bug tracker. |
| 51 |
. |
| 52 |
Thanks to every one who chipped in with suggestions. |
| 53 |
. |
| 54 |
Corbin |
Archives