allow CONFIG_TRIM_UNUSED_KSYM to be enabled when build as a builtin-module

[twoertwein]

Motivation and Context

When ZFS is built into the kernel, ZFS seems to work perfectly fine even when unused kernel symbols are removed.

Description

This patch allows CONFIG_TRIM_UNUSED_KSYM when enable_linux_builtin is used.

How Has This Been Tested?

Compiled 0.8.0 with this patch and used zfs (Archlinux, linux-lts).

Types of changes

• Bug fix (non-breaking change which fixes an issue)
• New feature (non-breaking change which adds functionality)
• Performance enhancement (non-breaking change which improves efficiency)
• Code cleanup (non-breaking change which makes code smaller or more readable)
• Breaking change (fix or feature that would cause existing functionality to change)
• Documentation (a change to man pages or other documentation)

Checklist:

• My code follows the ZFS on Linux code style requirements.
• I have updated the documentation accordingly.
• I have read the contributing document.
• I have added tests to cover my changes.
• All new and existing tests passed.
• All commit messages are properly formatted and contain Signed-off-by.

[codecov]

Codecov Report

Merging #8820 into master will increase coverage by 0.04%.
The diff coverage is n/a.

@@            Coverage Diff             @@
##           master    #8820      +/-   ##
==========================================
+ Coverage   78.73%   78.77%   +0.04%     
==========================================
  Files         382      382              
  Lines      117812   117812              
==========================================
+ Hits        92758    92812      +54     
+ Misses      25054    25000      -54

Flag	Coverage Δ
#kernel	79.37% <ø> (+0.03%)	⬆️
#user	67.58% <ø> (+0.32%)	⬆️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 841a7a9...dc9b7df. Read the comment docs.

[behlendorf]

Thanks, adding a buildin exclusion makes good sense. By definition when building with the rest of the kernel nothing should have been trimmed yet. Could you just verify from the config.log that the HAVE_PDE_DATA check in config/kernel-pde-data.m4 is working correctly. According to the commit which introduce this check 4b9dddf it's the reason this was required.

[twoertwein]

I assume this is the relevant part of config.log:

configure:36799: checking whether PDE_DATA() is available
configure:36833: cp conftest.c conftest.h build && make modules -C /home/twoertwein/AUR/linux-lts/src/linux-4.19.45 EXTRA_CFLAGS=-Werror -Wframe-larger-than=4096   M=/home/twoertwein/AUR/linux-lts/src/zfs-zfs-0.8.0/build modpost=true
configure:36836: $? = 0
configure:36838: test -s build/conftest.o
configure:36841: $? = 0
configure:36893: result: yes

it seems to be happy :)

[twoertwein]

it would probably be good to extend the automated built-in test-case (simply duplicate it but have CONFIG_TRIM_UNUSED_KSYM=y?) to check whether it introduces some unexpected behavior? "buildbot/Kernel.org Built-in x86_64 (BUILD)" seems to fail due to a git issue?

[behlendorf]

Looks good. Yes, I think it would be good to follow up with a tweak to the CI to define CONFIG_TRIM_UNUSED_KSYM=y in our custom builtin kernel builds. But let's delay that for a short while after merging this to prevent an unnecessary CI failures due to it.