microblaze: apply binutils patches from meta-xilinx repository #10
Conversation
alpsayin
force-pushed
the
zephyr-binutils-2_38-microblaze-recipes-28082023
branch
2 times, most recently
from
6a51ef5
to
1966eda
Compare
alpsayin
changed the title
alpsayin
force-pushed
the
zephyr-binutils-2_38-microblaze-recipes-28082023
branch
from
1966eda
to
8085bc5
Compare
alpsayin
force-pushed
the
zephyr-binutils-2_38-microblaze-recipes-28082023
branch
from
8085bc5
to
71acd97
Compare
|
@stephanosio resolved the marked issues and some more that I noticed irrelevant to microblaze or binutils. Zephyr(wout/ picolibc tests) and picolibc's own tests passed. |
bfd/elflink.c
Outdated
| @@ -6580,7 +6580,6 @@ elf_gc_sweep_symbol (struct elf_link_hash_entry *h, void *data) | |||
|
|
|||
| inf = (struct elf_gc_sweep_symbol_info *) data; | |||
| (*inf->hide_symbol) (inf->info, h, true); | |||
| h->def_regular = 0; | |||
There was a problem hiding this comment.
doesn't look microblaze related; to be inspected
There was a problem hiding this comment.
upstream change to garbage collection sweep causes mb regression
There was a problem hiding this comment.
Sounds like it fixes an issue. Keeping it in.
There was a problem hiding this comment.
doesn't look microblaze related; to be inspected
That was exactly my concern -- I was going to leave a comment about it then got interrupted by something else and it went off my radar ... (sorry about that).
My concern is its potential effect on other architectures -- will it not cause any regressions on other architectures relying on def_regular to be set to 0 by this function?
Unless we can prove it will not cause any regressions for other architectures, could we make this specific to Microblaze only? (maybe make the function accept bfd param and add this under if (bfd_get_arch(abfd) == bfd_arch_microblaze)?)
There was a problem hiding this comment.
I think I'll undo the patch for the time being because also from the original commit message, this seems very non-zephyr+embedded-linux related. If this gets upstreamed, and then if it finds its way down here through upstream binutils, then great, but otherwise a very edge case for an embedded application (also considering C++ is (should be) an edge case for an embedded application).
The visible issue with this change is when running a c++ application
in Petalinux which links libstdc++.so for exception handling it segfaults
on execution.
This does not occur if static linking libstdc++.a, so its during the
relocations for a shared lib with garbage collection this occurs
I'll now take a look at the others too.
alpsayin
force-pushed
the
zephyr-binutils-2_38-microblaze-recipes-28082023
branch
from
ee6f0da
to
04e5b36
Compare
This patchset fixes many known issues on gnu-toolchain for microblaze. But mainly the atomic-cas and binutils incorrect relocation issues. Patches obtained from https://github.com/xilinx/meta-xilinx/ repository, from meta-microblaze/recipes-devtools/binutils path. Below is a list of patches applied and squashed. Some patches that were already applied were reverted and re-applied to avoid conflicts. 0001-Add-wdc.ext.clear-and-wdc.ext.flush-insns.patch 0002-Add-mlittle-endian-and-mbig-endian-flags.patch 0003-Disable-the-warning-message-for-eh_frame_hdr.patch 0004-LOCAL-Fix-relaxation-of-assembler-resolved-reference.patch 0006-Fix-bug-in-TLSTPREL-Relocation.patch 0007-Added-Address-extension-instructions.patch 0008-Add-new-bit-field-instructions.patch 0009-Patch-Microblaze-fixed-bug-in-GCC-so-that-It-will-su.patch 0010-fixing-the-constant-range-check-issue.patch 0011-Patch-Microblaze-Compiler-will-give-error-messages-i.patch 0012-Patch-MicroBlaze-initial-support-for-MicroBlaze-64-b.patch 0013-Patch-Microblaze-negl-instruction-is-overriding-rsub.patch 0014-Added-relocations-for-MB-X.patch 0015-Fixed-MB-x-relocation-issues.patch 0016-Fixing-the-branch-related-issues.patch 0017-Fixed-address-computation-issues-with-64bit-address.patch 0018-Patch-MicroBlaze-Adding-new-relocation-to-support-64.patch 0019-fixing-the-.bss-relocation-issue.patch 0020-Fixed-the-bug-in-the-R_MICROBLAZE_64_NONE-relocation.patch 0021-Revert-ld-Remove-unused-expression-state.patch 0022-fixing-the-long-long-long-mingw-toolchain-issue.patch 0023-Added-support-to-new-arithmetic-single-register-inst.patch 0024-Patch-MicroBlaze-double-imml-generation-for-64-bit-v.patch 0025-Fixed-bug-in-generation-of-IMML-instruction-for-the.patch 0026-Patch-MicroBlaze-m64-This-patch-will-remove-imml-0-a.patch 0027-Patch-MicroBlaze-improper-address-mapping-of-PROVIDE.patch 0028-Patch-microblaze-Changing-the-long-to-long-long-as-i.patch 0029-gas-revert-moving-of-md_pseudo_table-from-const.patch 0030-ld-emulparams-elf64microblaze-Fix-emulation-generati.patch 0031-Patch-MicroBlaze-Invalid-data-offsets-pointer-after-.patch 0032-Patch-MicroBlaze-Double-free-with-ld-no-keep-memory.patch 0033-Patch-MB-MB-binutils-Upstream-port-issues.patch 0034-Patch-MicroBlaze-By-default-the-linker-will-generate.patch Signed-off-by: Alp Sayin <alpsayin@gmail.com>
This is because I don't intend to offer Zephyr support for microblaze64 yet but also because the toolchain is buggy to the extent that it fails to build. Signed-off-by: Alp Sayin <alpsayin@gmail.com>
alpsayin
force-pushed
the
zephyr-binutils-2_38-microblaze-recipes-28082023
branch
from
04e5b36
to
324e5d8
Compare
|
Hi @stephanosio , would you be happy to do a re-review? (when you're available) |
microblaze: apply binutils patches from meta-xilinx repository
This patchset fixes many known issues on gnu-toolchain for microblaze.
But mainly the atomic-cas and binutils incorrect relocation issues.
Patches obtained from https://github.com/xilinx/meta-xilinx/ repository,
from meta-microblaze/recipes-devtools/binutils path.
Some patches that were already applied were reverted and re-applied to avoid conflicts.
Finally, I've given up and disabled microblaze64 toolchain build. meta-xilinx patches somehow manage to break the same toolchain that they introduced. I have no intention to support microblaze64 yet and this saves a few CI cycles.