[v9,0/3] riscv: Add remaining module relocations and tests

Message

Charlie Jenkins Nov. 1, 2023, 6:32 p.m. UTC

A handful of module relocations were missing, this patch includes the
remaining ones. I also wrote some test cases to ensure that module
loading works properly. Some relocations cannot be supported in the
kernel, these include the ones that rely on thread local storage and
dynamic linking.

This patch also overhauls the implementation of ADD/SUB/SET/ULEB128
relocations to handle overflow. "Overflow" is different for ULEB128
since it is a variable-length encoding that the compiler can be expected
to generate enough space for. Instead of overflowing, ULEB128 will
expand into the next 8-bit segment of the location.

A psABI proposal [1] was merged that mandates that SET_ULEB128 and
SUB_ULEB128 are paired, however the discussion following the merging of
the pull request revealed that while the pull request was valid, it
would be better for linkers to properly handle this overflow. This patch
proactively implements this methodology for future compatibility.

This can be tested by enabling KUNIT, RUNTIME_KERNEL_TESTING_MENU, and
RISCV_MODULE_LINKING_KUNIT.

[1] https://github.com/riscv-non-isa/riscv-elf-psabi-doc/pull/403

Signed-off-by: Charlie Jenkins <[e-mail]>
---
Changes in v9:
- Use uintptr_t where applicable (Emil)
- Split up insn_rmw changes into correct patch (Emil)
- Make reloc_handlers const and simplify struct (Emil)
- Link to v8: https://lore.kernel.org/r/[e-mail]

Changes in v8:
- Fix up riscv_insn_rmw to work with big-endian systems (Andreas/Emil)
- Improve readability of bitops throughout module.c (Emil)
- Consolidate all relocation handler functions into a single struct so
  array doesn't need to be duplicated 3 times
- Use .space assember op instead of .rept
- Fix up name of uleb128 testcase
- Define prototypes of some new functions to conform to -Wmissing-prototypes
- Remove unused variable "first"
- Link to v7: https://lore.kernel.org/r/[e-mail]

Changes in v7:
- Overhaul ADD/SUB/SET/ULEB128 relocations
- Fix ULEB128 so it produces correct values when more than 1 byte is
  needed
- Link to v6: https://lore.kernel.org/r/[e-mail]

Changes in v6:
- Use (void *) instead of (u32 *) for handler type
- Constrain ULEB128 to be consecutive relocations
- Link to v5: https://lore.kernel.org/r/[e-mail]

Changes in v5:
- Brought in patch by Emil and fixed it up to force little endian
- Fixed up issues with apply_r_riscv_32_pcrel_rela and
  apply_r_riscv_plt32_rela (Samuel)
- Added u8 cast in apply_r_riscv_sub6_rela (Andreas) 
- Link to v4: https://lore.kernel.org/r/[e-mail]

Changes in v4:
- Complete removal of R_RISCV_RVC_LUI
- Fix bug in R_RISCV_SUB6 linking
- Only build ULEB128 tests if supported by toolchain
- Link to v3: https://lore.kernel.org/r/[e-mail]

Changes in v3:
- Add prototypes to test_module_linking_main as recommended by intel
  zero day bot
- Improve efficiency of ULEB128 pair matching
- Link to v2: https://lore.kernel.org/r/[e-mail]

Changes in v2:
- Added ULEB128 relocations
- Link to v1: https://lore.kernel.org/r/[e-mail]

---
Charlie Jenkins (2):
      riscv: Add remaining module relocations
      riscv: Add tests for riscv module loading

Emil Renner Berthing (1):
      riscv: Avoid unaligned access when relocating modules

 arch/riscv/Kconfig.debug                           |   1 +
 arch/riscv/include/uapi/asm/elf.h                  |   5 +-
 arch/riscv/kernel/Makefile                         |   1 +
 arch/riscv/kernel/module.c                         | 603 +++++++++++++++++----
 arch/riscv/kernel/tests/Kconfig.debug              |  35 ++
 arch/riscv/kernel/tests/Makefile                   |   1 +
 arch/riscv/kernel/tests/module_test/Makefile       |  15 +
 .../tests/module_test/test_module_linking_main.c   |  88 +++
 arch/riscv/kernel/tests/module_test/test_set16.S   |  23 +
 arch/riscv/kernel/tests/module_test/test_set32.S   |  20 +
 arch/riscv/kernel/tests/module_test/test_set6.S    |  23 +
 arch/riscv/kernel/tests/module_test/test_set8.S    |  23 +
 arch/riscv/kernel/tests/module_test/test_sub16.S   |  20 +
 arch/riscv/kernel/tests/module_test/test_sub32.S   |  20 +
 arch/riscv/kernel/tests/module_test/test_sub6.S    |  20 +
 arch/riscv/kernel/tests/module_test/test_sub64.S   |  25 +
 arch/riscv/kernel/tests/module_test/test_sub8.S    |  20 +
 arch/riscv/kernel/tests/module_test/test_uleb128.S |  31 ++
 18 files changed, 869 insertions(+), 105 deletions(-)
---
base-commit: 3bcce01fcbcd868b8cf3a5632fde283e122d7213
change-id: 20230908-module_relocations-f63ced651bd7

Comments

[e-mail] Nov. 8, 2023, 3:10 p.m. UTC | #1

Hello:

This series was applied to riscv/linux.git (for-next)
by Palmer Dabbelt <[e-mail]>:

On Wed, 01 Nov 2023 11:32:58 -0700 you wrote:
> A handful of module relocations were missing, this patch includes the
> remaining ones. I also wrote some test cases to ensure that module
> loading works properly. Some relocations cannot be supported in the
> kernel, these include the ones that rely on thread local storage and
> dynamic linking.
> 
> This patch also overhauls the implementation of ADD/SUB/SET/ULEB128
> relocations to handle overflow. "Overflow" is different for ULEB128
> since it is a variable-length encoding that the compiler can be expected
> to generate enough space for. Instead of overflowing, ULEB128 will
> expand into the next 8-bit segment of the location.
> 
> [...]

Here is the summary with links:
  - [v9,1/3] riscv: Avoid unaligned access when relocating modules
    https://git.kernel.org/riscv/c/8cbe0accc4a6
  - [v9,2/3] riscv: Add remaining module relocations
    https://git.kernel.org/riscv/c/8fd6c5142395
  - [v9,3/3] riscv: Add tests for riscv module loading
    https://git.kernel.org/riscv/c/af71bc194916

You are awesome, thank you!