BTF mismatch can occur for a separately-built module even when the ABI
is otherwise compatible and nothing else would prevent successfully
loading. Add a new config to control how mismatches are handled. By
default, preserve the current behavior of refusing to load the
module. If MODULE_ALLOW_BTF_MISMATCH is enabled, load the module but
ignore its BTF information.
Suggested-by: Yonghong Song <[email protected]>
Suggested-by: Michal Suchánek <[email protected]>
Signed-off-by: Connor O'Brien <[email protected]>
---
Hello,
In the discussion regarding BTF compatibility & modules, there seemed
to be broad agreement that an option to ignore mismatches would be
reasonable. Currently the only option for handling this problem seems
to be to disable BTF entirely, so this would at least be an
incremental improvement.
Thanks,
Connor
kernel/bpf/btf.c | 3 ++-
lib/Kconfig.debug | 10 ++++++++++
2 files changed, 12 insertions(+), 1 deletion(-)
diff --git a/kernel/bpf/btf.c b/kernel/bpf/btf.c
index 11740b300de9..1a21f24105b3 100644
--- a/kernel/bpf/btf.c
+++ b/kernel/bpf/btf.c
@@ -6397,7 +6397,8 @@ static int btf_module_notify(struct notifier_block *nb, unsigned long op,
pr_warn("failed to validate module [%s] BTF: %ld\n",
mod->name, PTR_ERR(btf));
kfree(btf_mod);
- err = PTR_ERR(btf);
+ if (!IS_ENABLED(CONFIG_MODULE_ALLOW_BTF_MISMATCH))
+ err = PTR_ERR(btf);
goto out;
}
err = btf_alloc_id(btf);
diff --git a/lib/Kconfig.debug b/lib/Kconfig.debug
index 1555da672275..ff857bb7d633 100644
--- a/lib/Kconfig.debug
+++ b/lib/Kconfig.debug
@@ -339,6 +339,16 @@ config DEBUG_INFO_BTF_MODULES
help
Generate compact split BTF type information for kernel modules.
+config MODULE_ALLOW_BTF_MISMATCH
+ bool "Allow loading modules with non-matching BTF type info"
+ depends on DEBUG_INFO_BTF_MODULES
+ help
+ For modules whose split BTF does not match vmlinux, load without
+ BTF rather than refusing to load. The default behavior with
+ module BTF enabled is to reject modules with such mismatches;
+ this option will still load module BTF where possible but ignore
+ it when a mismatch is found.
+
config GDB_SCRIPTS
bool "Provide GDB scripts for kernel debugging"
help
--
2.35.1.473.g83b2b277ed-goog
> On Feb 22, 2022, at 8:34 PM, Shung-Hsi Yu <[email protected]> wrote:
>
> On Wed, Feb 23, 2022 at 01:28:14AM +0000, Connor O'Brien wrote:
>> BTF mismatch can occur for a separately-built module even when the ABI
>> is otherwise compatible and nothing else would prevent successfully
>> loading. Add a new config to control how mismatches are handled. By
>> default, preserve the current behavior of refusing to load the
>> module. If MODULE_ALLOW_BTF_MISMATCH is enabled, load the module but
>> ignore its BTF information.
>>
>> Suggested-by: Yonghong Song <[email protected]>
>> Suggested-by: Michal Suchánek <[email protected]>
>> Signed-off-by: Connor O'Brien <[email protected]>
>
> Maybe reference the discussion thread as well?
>
> Link: https://lore.kernel.org/bpf/CAADnVQJ+OVPnBz8z3vNu8gKXX42jCUqfuvhWAyCQDu8N_yqqwQ@mail.gmail.com/
>
> Otherwise
>
> Acked-by: Shung-Hsi Yu <[email protected]>
Acked-by: Song Liu <[email protected]>
On Wed, Feb 23, 2022 at 01:28:14AM +0000, Connor O'Brien wrote:
> BTF mismatch can occur for a separately-built module even when the ABI
> is otherwise compatible and nothing else would prevent successfully
> loading. Add a new config to control how mismatches are handled. By
> default, preserve the current behavior of refusing to load the
> module. If MODULE_ALLOW_BTF_MISMATCH is enabled, load the module but
> ignore its BTF information.
>
> Suggested-by: Yonghong Song <[email protected]>
> Suggested-by: Michal Suchánek <[email protected]>
> Signed-off-by: Connor O'Brien <[email protected]>
Maybe reference the discussion thread as well?
Link: https://lore.kernel.org/bpf/CAADnVQJ+OVPnBz8z3vNu8gKXX42jCUqfuvhWAyCQDu8N_yqqwQ@mail.gmail.com/
Otherwise
Acked-by: Shung-Hsi Yu <[email protected]>
> ---
> Hello,
>
> In the discussion regarding BTF compatibility & modules, there seemed
> to be broad agreement that an option to ignore mismatches would be
> reasonable. Currently the only option for handling this problem seems
> to be to disable BTF entirely, so this would at least be an
> incremental improvement.
>
> Thanks,
> Connor
On 2/23/22 5:34 AM, Shung-Hsi Yu wrote:
> On Wed, Feb 23, 2022 at 01:28:14AM +0000, Connor O'Brien wrote:
>> BTF mismatch can occur for a separately-built module even when the ABI
>> is otherwise compatible and nothing else would prevent successfully
>> loading. Add a new config to control how mismatches are handled. By
>> default, preserve the current behavior of refusing to load the
>> module. If MODULE_ALLOW_BTF_MISMATCH is enabled, load the module but
>> ignore its BTF information.
>>
>> Suggested-by: Yonghong Song <[email protected]>
>> Suggested-by: Michal Suchánek <[email protected]>
>> Signed-off-by: Connor O'Brien <[email protected]>
>
> Maybe reference the discussion thread as well?
>
> Link: https://lore.kernel.org/bpf/CAADnVQJ+OVPnBz8z3vNu8gKXX42jCUqfuvhWAyCQDu8N_yqqwQ@mail.gmail.com/
LGTM, and added above into the commit log while applying. Thanks everyone!
Hello:
This patch was applied to bpf/bpf-next.git (master)
by Daniel Borkmann <[email protected]>:
On Wed, 23 Feb 2022 01:28:14 +0000 you wrote:
> BTF mismatch can occur for a separately-built module even when the ABI
> is otherwise compatible and nothing else would prevent successfully
> loading. Add a new config to control how mismatches are handled. By
> default, preserve the current behavior of refusing to load the
> module. If MODULE_ALLOW_BTF_MISMATCH is enabled, load the module but
> ignore its BTF information.
>
> [...]
Here is the summary with links:
- [bpf-next] bpf: add config to allow loading modules with BTF mismatches
https://git.kernel.org/bpf/bpf-next/c/5e214f2e43e4
You are awesome, thank you!
--
Deet-doot-dot, I am a bot.
https://korg.docs.kernel.org/patchwork/pwbot.html