Repository navigation
Consider removing the initcall logic from module.rs, use codegen instead #77
Description
Currently we duplicate a bit of logic on the Rust side for our initcall handling.
Consider performing codegen instead to be able to reuse the C macros instead. This is basically the approach we took in rust/exports.c too. Even if may look ugly, it is best to avoid redoing parts of the C infrastructure.
This would also remove our need for the unstable global_asm!.
Activity
kavirajk commentedon Mar 21, 2021
Can I work on this? :)
ojeda commentedon Mar 21, 2021
Sure!
Please note it is quite an open-ended issue though, and how to do it best is not clear (I say this because sometimes with this kind of issue one may do a lot of work to then find a given approach was not a good idea to begin with. It is also why I wrote "Consider" in the title).
kavirajk commentedon Mar 24, 2021
@ojeda Thanks for the heads up :)
Checking my understanding.
The goal is actually to replace handwritten
init_module,cleanup_moduleand the way we useinitcallwith existing corresponding c's functions correct? and the suggestion is to generate these FFI via bindgen correct?I have few questions: Hope you can clarify.
I see from the code that those function eventually calls
__init()and__exit()functions, which just set__MODvariable toSome(m)andNonerespectively. I don't understand what is this __MOD and where is it comes from? I grepped through the code and couldn't figure out.I don't quite understand how do we even replace
_initcallfunction, it use bare asm.Thanks in advance for your time :)
ojeda commentedon Mar 25, 2021
Not exactly! The idea is to generate a C source file on the fly that contains the invocations of the
MODULE_*macros from themodule!call. Then, compile that C file and link it with the module automatically.__MODrepresents the module (it is anOptionof the type that the user specified inmodule!) and is generated inrust/module.rs:linux/rust/module.rs
Line 584 in 4f0e492
dmaengine: xilinx: dpdma: initialize registers before request_irq
cifs: fix use-after-free on the link name
mm: kmem: properly initialize local objcg variable in current_obj_cgr…
bpf: Fix a few selftest failures due to llvm18 change
ocfs2: check tl->count of truncate log inode in ocfs2_get_truncate_lo…
fs/dax: Fix "don't skip locked entries when scanning entries"
ocfs2: check tl->count of truncate log inode in ocfs2_get_truncate_lo…