summaryrefslogtreecommitdiffstats
path: root/meta-microblaze
Commit message (Collapse)AuthorAgeFilesLines
...
* gdb: Fix error in inline_frame_snifferMark Hatle2022-07-282-0/+62
| | | | | | | | | Depth: Total number of inline functions [refer inline-frame.c] state->skipped_frames : Number of inline functions skipped. the current unwind_pc is causing an issue when we try to step into inline functions[Depth is becoming 0]. It’s incrementing pc by 8 even with si instruction. Signed-off-by: Mark Hatle <mark.hatle@amd.com>
* meta-microblaze: gdb is now availableMark Hatle2022-07-281-2/+0
| | | | Signed-off-by: Mark Hatle <mark.hatle@amd.com>
* meta-microbalze: gdb: Enable gdb and gdbserver on microblazeMark Hatle2022-07-2814-985/+1301
| | | | | | | Replace existing integration with new work that enables target gdb and gdbserver. Signed-off-by: Mark Hatle <mark.hatle@amd.com>
* ltp: Disable ltp gdb dependency only on MicroblazeMark Hatle2022-03-281-0/+2
| | | | | | | On-target gdb does not work on microblaze, disable the dependency. This may affect certain ltp test cases. Signed-off-by: Mark Hatle <mark.hatle@xilinx.com>
* gdb: Disable on-target GDB for microblazeMark Hatle2022-03-281-0/+7
| | | | | | | | | | | | | Failure: ../../gdb-10.2/bfd/cpu-microblaze.c:75:1: warning: missing initializer for field 'max_reloc_offset_into_insn' of 'bfd_arch_info_type' {aka 'const struct bfd_arch_info'} [-Wmissing-field-initializers] | 75 | }, | | ^ | In file included from ../../gdb-10.2/bfd/cpu-microblaze.c:23: | ./bfd.h:1998:14: note: 'max_reloc_offset_into_insn' declared here | 1998 | signed int max_reloc_offset_into_insn; | | ^~~~~~~~~~~~~~~~~~~~~~~~~~ Signed-off-by: Mark Hatle <mark.hatle@xilinx.com>
* meta-microblaze: Move to glibc backportMark Hatle2022-03-283-51/+35
| | | | | | Move to a backport of the official fix for glibc BZ #28883. Signed-off-by: Mark Hatle <mark.hatle@xilinx.com>
* meta-microblaze: Set OLDEST_KERNEL version is 3.15Mark Hatle2022-02-121-0/+2
| | | | | | | Version 3.15 is required to enable pselect instead of (new)select syscall when using the select function. Signed-off-by: Mark Hatle <mark.hatle@xilinx.com>
* meta-microblaze: Fix glibc issues related to select/pselectMark Hatle2022-02-124-0/+122
| | | | | | | | | | | | | *) Select syscall is not implemented on microblaze, newselect should be used instead, but commit 4c3df0eba5e8fe98f0de917ade9b2ebba6951c5f from glibc removed this behavior. Presumably by accident. *) When the code path that doesn't use select, but instead uses pselect32 is used, microblaze provides it's own implementation but this implementation is only available when ASSUME_PSELECT is disabled. We need to fallback to the default Linux implementation. Signed-off-by: Mark Hatle <mark.hatle@xilinx.com>
* microblaze binutils: Integrate community binutils contributionsMark Hatle2022-02-044-0/+388
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | Three bug fixes suggested by community member dednev@rambler.ru: *elf*-microblaze-relaxation.patch short description: The roots of invalid data offsets (pointer) after relaxation are in internal symbol table cache handling. If symbol table was cached while section processing for garbage collection, the elf_tdata (abfd)->symtab_hdr pointer is not NULL, but (Elf_Internal_Sym *) symtab_hdr->contents array contains data for local symbols only. But in the repeated relocation processing there is no check whenever the symbol index is local (ELF32_R_SYM (irelscan->r_info) < symtab_hdr->sh_info) or global. This leads to the buffer overflow (global symbol index is out of bounds) and following invalid relocation processing, especially in extra relocation data modifications (f.e. irelscan->r_addend could be decreased by arbitrary values). *elf*-microblaze-no-keep-memory.patch short description: There is incorrect newly allocated buffers handling if internal memory caching is disabled (without --no-keep-memory ld switch). This leads to possible double free(...) calls for such data and results in unexpected ld termination with abort(). I've fixed this issue by implementing similar to the elf32-avr.c data buffer handling. *elf*-imm-check-for-relaxation.patch short description: I've reverted Xilinx path to the original GNU binutils bfd code for microblaze, because this patch incorrectly leaves some "imm -1" instructions in resulting binary file after relaxation and increases executable size. If you look at the IMM instruction argument processing in binutils/gas/config/tc-microblaze.c you will see, that gas checks are similar to the original bfd microblaze_elf_relax_section(...) Signed-off-by: Mark Hatle <mark.hatle@xilinx.com>
* Remove hardknott from LAYERSERIES_COMPATJohn Toomey2022-02-041-1/+1
| | | | | | | We do not maintain support for hardknott so it is being removed Signed-off-by: John Toomey <jtoomey@xilinx.com> Signed-off-by: Mark Hatle <mark.hatle@xilinx.com>
* Remove gatesgarth from LAYERSERIES_COMPATJohn Toomey2022-02-041-1/+1
| | | | | | | | We do not build, test or support gatesgarth in any of our layers so it makes sense to remove this for the various layer.conf files Signed-off-by: John Toomey <jtoomey@xilinx.com> Signed-off-by: Mark Hatle <mark.hatle@xilinx.com>
* Add missing README.mdJohn Toomey2022-02-041-0/+28
| | | | | Signed-off-by: John Toomey <jtoomey@xilinx.com> Signed-off-by: Mark Hatle <mark.hatle@xilinx.com>
* Initial restructure/split of meta-xilinx-bspMark Hatle2022-01-141-1/+1
| | | | | | | | | | Create a new meta-xilinx-core, move core functionality to the core, keeping board specific files in the bsp layer. zynqmp-generic changed from require <board> to include, so if meta-xilinx-bsp is not available it will not fail. Signed-off-by: Mark Hatle <mark.hatle@xilinx.com>
* microblaze systemd: Add additional missing architecture supportMark Hatle2022-01-146-1/+868
| | | | | | | | | | | | | | Rename a mispelled patch name Add a patch that defines the microblaze syscalls. Disable stack-protection, as it is not supported on Microblaze at this time. Add a patch to resolve an undefined gcc internal, that is not enabled on microblaze. Signed-off-by: Mark Hatle <mark.hatle@xilinx.com>
* microblaze systemd: Move initial patch from meta-petalinuxMark Hatle2022-01-142-0/+61
| | | | Signed-off-by: Mark Hatle <mark.hatle@xilinx.com>
* microblaze m4/diffutils/grep: Add stack direction info to m4 macrosMark Hatle2022-01-146-0/+45
| | | | | | | | | | | | These tools require knowledge of the stack direction, since we are cross compiling they are unable to detect the stack growth direction so it must be defined in the m4 macro. It turns out the direction was already set, but defined as 'microblaze' instead of 'microblaze*' which caused problems on little endian. Signed-off-by: Mark Hatle <mark.hatle@xilinx.com>
* microblaze libxcrypt: Fall back to older style symver supportMark Hatle2022-01-142-0/+19
| | | | | | | | | | | | | Using symver related functions results in: | ../git/lib/crypt-static.c:32:1: error: symver is only supported on ELF platforms | 32 | SYMVER_crypt; | | ^~~~~~~~~~~~ Microblaze does not support these functions at this time, so fall back to the older approach to resolve this issue. Signed-off-by: Mark Hatle <mark.hatle@xilinx.com>
* newlib: Update to honister versionMark Hatle2022-01-1413-72/+50
| | | | Signed-off-by: Mark Hatle <mark.hatle@xilinx.com>
* gdb: Upgrade to the honister versionMark Hatle2022-01-1468-6140/+2579
| | | | | | | | Move from pinned gdb version 9.2 to the honister version. Note: gdb/gdbserver for the target is known to not work yet. Signed-off-by: Mark Hatle <mark.hatle@xilinx.com>
* gcc: Upgrade to honister (11.2) versionMark Hatle2022-01-1457-710/+465
| | | | Signed-off-by: Mark Hatle <mark.hatle@xilinx.com>
* binutils: Upgraade to honister versionMark Hatle2022-01-1459-5446/+3912
| | | | Signed-off-by: Mark Hatle <mark.hatle@xilinx.com>
* gcc: Move everything to gcc-11 directoryMark Hatle2022-01-142-1/+1
| | | | Signed-off-by: Mark Hatle <mark.hatle@xilinx.com>
* Revert "binutils: Import gatesgarth version, converted to new variable syntax"Mark Hatle2022-01-1429-25355/+0
| | | | | | This reverts commit 294443d89dd53435fc0ccc6113c7ec5aff326612. Signed-off-by: Mark Hatle <mark.hatle@xilinx.com>
* Revert "binutils: convert away from bbappend"Mark Hatle2022-01-145-4/+15
| | | | | | This reverts commit 7f12056ccc3a978e9b69c380bffc3c04ee20520e. Signed-off-by: Mark Hatle <mark.hatle@xilinx.com>
* Revert "gcc: Import gatesgarth version, converted to new variable syntax"Mark Hatle2022-01-1469-7082/+0
| | | | | | This reverts commit 2c819946b5258291d1725dabb3f5180d36971bd6. Signed-off-by: Mark Hatle <mark.hatle@xilinx.com>
* Revert "gcc: convert away from bbappend"Mark Hatle2022-01-1459-2/+3
| | | | | | This reverts commit b56b1562c129fd12b486fd6666315afd4644ce6a. Signed-off-by: Mark Hatle <mark.hatle@xilinx.com>
* Convert to new override syntaxSai Hari Chandana Kalluri2021-11-023-4/+4
| | | | | | | | | | This is the result of automated script (0.9.0) conversion: oe-core/scripts/contrib/convert-overrides.py . converting the metadata to use : as the override character instead of _. Signed-off-by: Sai Hari Chandana Kalluri <chandana.kalluri@xilinx.com>
* newlib: Enable --disable-newlib-reent-check-verifyMark Hatle2021-11-021-0/+4
| | | | | | | | | | | | | In the newlib-3.3 release reent-check is enabled by default. The reent-check is needed to ensure memory is not NULL. See: https://sourceware.org/git/?p=newlib-cygwin.git;a=commit;h=8b39f7406c7b819bc45a83e9c94531221a6b3b34 We need to disable this check on microblaze, as iostream library test cases fail with an assert error. Signed-off-by: Mark Hatle <mark.hatle@xilinx.com>
* zstd: Add libatomic for linkingMark Hatle2021-10-251-0/+2
| | | | Signed-off-by: Mark Hatle <mark.hatle@xilinx.com>
* gcc: convert away from bbappendMark Hatle2021-10-2559-3/+2
| | | | | | Intergrade the microblaze items with the backported toolchain. Signed-off-by: Mark Hatle <mark.hatle@xilinx.com>
* gcc: Import gatesgarth version, converted to new variable syntaxMark Hatle2021-10-2569-0/+7082
| | | | Signed-off-by: Mark Hatle <mark.hatle@xilinx.com>
* binutils: convert away from bbappendMark Hatle2021-10-255-15/+4
| | | | Signed-off-by: Mark Hatle <mark.hatle@xilinx.com>
* binutils: Import gatesgarth version, converted to new variable syntaxMark Hatle2021-10-2529-0/+25355
| | | | Signed-off-by: Mark Hatle <mark.hatle@xilinx.com>
* Convert to new override syntaxSai Hari Chandana Kalluri2021-09-2818-38/+38
| | | | | | | | | | This is the result of automated script (0.9.0) conversion: oe-core/scripts/contrib/convert-overrides.py . converting the metadata to use ":" as the override character instead of "_". Signed-off-by: Sai Hari Chandana Kalluri <chandana.kalluri@xilinx.com>
* meta-microblaze: Fix GCC issue where additional loops were addedMark Hatle2021-09-282-0/+58
| | | | | | | | Due to latest changes in GCC-10.2 MB64 perforamance has reduced We have added zero_extended instructions to get rid of left shift and right shift loops Signed-off-by: Mark Hatle <mark.hatle@xilinx.com>
* Revert "gcc: Fix performance degradation on Microblaze 64"Mark Hatle2021-09-282-52/+0
| | | | | | This reverts commit 1bb08d385a4dbd1a3bf07a3917ea32f4be91ffac. Signed-off-by: Mark Hatle <mark.hatle@xilinx.com>
* gcc: Fix performance degradation on Microblaze 64Mark Hatle2021-09-282-0/+52
| | | | | | | | | | | | Issue: TSR-974519 Added zero_extendqidi2 and zero_extendhidi2 patterns in GCC 10.2,commit 0a237a9: Do not propagate results from inner REGS to paradoxical SUBREGs and this commit is causing the MB 64 compiler to generate extra instructions. Signed-off-by: Mark Hatle <mark.hatle@xilinx.com>
* meta-microblaze: Upgrade gcc-source 10 -> 11Sai Hari Chandana Kalluri2021-07-1557-0/+0
| | | | Signed-off-by: Sai Hari Chandana Kalluri <chandana.kalluri@xilinx.com>
* Upgrade newlib, libgloss 3.3 -> 4.1Sai Hari Chandana Kalluri2021-07-152-0/+0
| | | | | | Upgrade newlib, libgloss recipes from 3.3 -> 4.1 Signed-off-by: Sai Hari Chandana Kalluri <chandana.kalluri@xilinx.com>
* layer.conf: Add honister hardknott LAYER COMPATSai Hari Chandana Kalluri2021-07-151-1/+1
| | | | Signed-off-by: Sai Hari Chandana Kalluri <chandana.kalluri@xilinx.com>
* gcc: Fix break_hander attribute handlingMark Hatle2021-07-142-0/+72
| | | | | | Issue: CR-1081780 Signed-off-by: Mark Hatle <mark.hatle@xilinx.com>
* binutis-gdb: Fix size of long, should be 64-bitMark Hatle2021-07-142-0/+28
| | | | Signed-off-by: Mark Hatle <mark.hatle@xilinx.com>
* binutils/gdb: Disable hunk from microblaze patchMark Hatle2021-07-142-2/+2
| | | | | | | | | | The disabled chunk triggers a compilation failure when trying to build newlib for microblaze64. ld: MicroBlaze architecture of input file `/tmp/ccHA6bsQ.o' is incompatible with MicroBlaze output ld: warning: cannot find entry symbol _start; defaulting to 0000000000000050 Signed-off-by: Mark Hatle <mark.hatle@xilinx.com>
* binutils/gdb: Fix various microblaze 64 issuesMark Hatle2021-07-146-13/+412
| | | | | | | Issue: CR-1089197 Issue: CR-1093443 Signed-off-by: Mark Hatle <mark.hatle@xilinx.com>
* layer.conf: compatible with gatesgarth onlyMark Hatle2021-07-141-2/+1
| | | | Signed-off-by: Mark Hatle <mark.hatle@xilinx.com>
* libgloss: Fix crt*.o files that are stripped, when they shouldn't beMark Hatle2021-07-141-0/+7
| | | | | | | | | | | Issue: CR-1062133 For some reason libgloss installs the crt*.o files with executable bit set, this causes the build system to strip these as executables. However, it makes it impossible to use the .o files to properly link applications later on. Signed-off-by: Mark Hatle <mark.hatle@xilinx.com>
* binutils: Workaround for Microblaze 64 linking failureMark Hatle2021-07-141-9/+0
| | | | | | | | The typecast is causing linking to not work properly, however it does cause a compilation warning to return that a typecast is likely needed. Signed-off-by: Mark Hatle <mark.hatle@xilinx.com> Signed-off-by: Mark Hatle <mark.hatle@kernel.crashing.org>
* meta-microblaze gcc-source: Fix barrel shift for versions under v10.0Mark Hatle2021-07-142-0/+50
| | | | | | | Issue reported by Daniel Mack via the meta-xilinx mailing list. Signed-off-by: Mark Hatle <mark.hatle@xilinx.com> Signed-off-by: Mark Hatle <mark.hatle@kernel.crashing.org>
* Revert "newlib/libgloss: YP has moved to version 4.1.0, adjust accordingly"Sai Hari Chandana Kalluri2021-07-142-0/+0
| | | | | | This reverts commit 96f122efe48f239f7b5df4025acd5c98a61828b3. Signed-off-by: Sai Hari Chandana Kalluri <chandana.kalluri@xilinx.com>
* newlib/libgloss: YP has moved to version 4.1.0, adjust accordinglyMark Hatle2021-02-032-0/+0
| | | | Signed-off-by: Mark Hatle <mark.hatle@kernel.crashing.org>
generated by cgit v1.2.3-54-g00ecf (git 2.39.0) at 2025-07-23 09:21:13 +0000