Skip to contentSkip to conversation

rust: Use `ffi::c_char` type in firmware abstraction `FwFunc`

PATCH Superseded Christian Schrefl 3 participants 14 messages Apr 12, 2025 – Apr 13, 2025

Subsystems (2)

v2 of 3 Superseded Latest is v3 →

lore
Share & tools
lore
#0 lore raw

PATCH v2 1/1 rust: Use `ffi::c_char` type in firmware abstraction `FwFunc`

The FwFunc struct contains an function with a char pointer argument, for which a *const u8 pointer was used. This is not really the "propper" type for this, so use a *const kernel::ffi::c_char pointer instad.

This has no real functionality changes, since kernel::ffi::c_char is a type alias to u8 anyways.

This used to cause problems on 6.13 when building for 32 bit arm (with my patches), since rust mapped c_char to i8 instead.
Build error for this case:

```
error[E0308]: mismatched types
  --> rust/kernel/firmware.rs:20:4
   |
20 |         Self(bindings::request_firmware)
   |         ---- ^^^^^^^^^^^^^^^^^^^^^^^^^^ expected fn pointer, found fn item
   |         |
   |         arguments to this function are incorrect
   |
   = note: expected fn pointer `unsafe extern "C" fn(_, *const i8, _) -> _`
                 found fn item `unsafe extern "C" fn(_, *const u8, _) -> _ {request_firmware}`
note: tuple struct defined here
  --> rust/kernel/firmware.rs:14:8
   |
14 | struct FwFunc(
   |        ^^^^^^

error[E0308]: mismatched types
  --> rust/kernel/firmware.rs:24:14
   |
24 |         Self(bindings::firmware_request_nowarn)
   |         ---- ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ expected fn pointer, found fn item
   |         |
   |         arguments to this function are incorrect
   |
   = note: expected fn pointer `unsafe extern "C" fn(_, *const i8, _) -> _`
                 found fn item `unsafe extern "C" fn(_, *const u8, _) -> _ {firmware_request_nowarn}`
note: tuple struct defined here
  --> rust/kernel/firmware.rs:14:8
   |
14 | struct FwFunc(
   |        ^^^^^^

error[E0308]: mismatched types
  --> rust/kernel/firmware.rs:64:45
   |
64 |         let ret = unsafe { func.0(pfw as _, name.as_char_ptr(), dev.as_raw()) };
   |                            ------           ^^^^^^^^^^^^^^^^^^ expected `*const i8`, found `*const u8`
   |                            |
   |                            arguments to this function are incorrect
   |
   = note: expected raw pointer `*const i8`
              found raw pointer `*const u8`

error: aborting due to 3 previous errors
```
Fixes de6582833db0 ("rust: add firmware abstractions")

---
This should probably be backported to stable, for people/distros
using Arm 32 patches on stable.
---
Changes in v2:
- Use kernel::ffi::c_char instead of core::ffi::c_char. (Danilo & Benno)
- Reword the commit message.
- Link to v1: https://lore.kernel.org/r/[email protected]

---
 rust/kernel/firmware.rs | 8 ++++++--
 1 file changed, 6 insertions(+), 2 deletions(-)
rust/kernel/firmware.rs +6 -2
index f04b058b09b2d2397e26344d0e055b3aa5061432..2494c96e105f3a28af74548d63a44464ba50eae3 100644
@@ -4,7 +4,7 @@
 //!
 //! C header: [`include/linux/firmware.h`](srctree/include/linux/firmware.h)
 
-use crate::{bindings, device::Device, error::Error, error::Result, str::CStr};
+use crate::{bindings, device::Device, error::Error, error::Result, ffi, str::CStr};
 use core::ptr::NonNull;
 
 /// # Invariants
@@ -12,7 +12,11 @@
 /// One of the following: `bindings::request_firmware`, `bindings::firmware_request_nowarn`,
 /// `bindings::firmware_request_platform`, `bindings::request_firmware_direct`.
 struct FwFunc(
-    unsafe extern "C" fn(*mut *const bindings::firmware, *const u8, *mut bindings::device) -> i32,
+    unsafe extern "C" fn(
+        *mut *const bindings::firmware,
+        *const ffi::c_char,
+        *mut bindings::device,
+    ) -> i32,
 );
 
 impl FwFunc {

---
base-commit: 0af2f6be1b4281385b618cb86ad946eded089ac8
change-id: 20250408-rust_arm_fix_fw_abstaction-4c3a89d75e29

Best regards,

Signature · 1 line
Christian Schrefl <[email protected]>
#1 lore raw
Quoted 3 lines · Christian Schrefl

"propper" type for this, so use a *const kernel::ffi::c_char pointer instad.

A couple typos -- I usually suggest using e.g. scripts/checkpatch.pl --codespell.

Quoted 2 lines

This should probably be backported to stable, for people/distros using Arm 32 patches on stable.

Up to the stable team -- I will add Cc: stable when I pick it up since it shouldn't hurt, but it does not change anything there for the supported arches, so they may or may not want to pick it.

Thanks!

Cheers,
Miguel

maintainer Acked-by
#2 lore raw
Quoted 9 lines · Miguel Ojeda
Quoted 3 lines · Christian Schrefl

"propper" type for this, so use a *const kernel::ffi::c_char pointer instad.

A couple typos -- I usually suggest using e.g. scripts/checkpatch.pl --codespell.

Quoted 6 lines
Quoted 2 lines

This should probably be backported to stable, for people/distros using Arm 32 patches on stable.

Up to the stable team -- I will add Cc: stable when I pick it up since it shouldn't hurt, but it does not change anything there for the supported arches, so they may or may not want to pick it.

Typically, firmware patches go through driver-core, but if you like I'm fine with you taking it through the Rust tree.

I don't think backporting it hurts, but it's also not really necessary. Like Miguel says, up to the stable team.

Just note that before v6.13, core::ffi is required, since kernel::ffi does not exist. The firmware abstractions were introduced with v6.11.

- Danilo

Acked-by
#3 lore raw
Quoted 5 lines · Danilo Krummrich

Typically, firmware patches go through driver-core, but if you like I'm fine with you taking it through the Rust tree.

Happy either way! If someone else takes it:

Quoted 2 lines

Just note that before v6.13, core::ffi is required, since kernel::ffi does not exist. The firmware abstractions were introduced with v6.11.

It exists nowadays in 6.12.y (I backported it with the big alloc backport series).

Cheers,
Miguel

#4 lore raw
In reply to Miguel Ojeda
Quoted 8 lines · Miguel Ojeda
Quoted 3 lines · Christian Schrefl

"propper" type for this, so use a *const kernel::ffi::c_char pointer instad.

A couple typos -- I usually suggest using e.g. scripts/checkpatch.pl --codespell.

Do you want me to send a v3 with that fixed or are you going to do that when applying.

(I've now configured my b4 check to add --codespell)

Quoted 12 lines
Quoted 3 lines

This should probably be backported to stable, for people/distros using Arm 32 patches on stable.

Up to the stable team -- I will add Cc: stable when I pick it up since it shouldn't hurt, but it does not change anything there for the supported arches, so they may or may not want to pick it.

Thanks!

Cheers,
Miguel

#5 lore raw
Quoted 3 lines · Christian Schrefl

Do you want me to send a v3 with that fixed or are you going to do that when applying.

No, no worries, but thanks :)

Cheers,
Miguel

maintainer
#6 lore raw
In reply to Christian Schrefl
Quoted 10 lines · Christian Schrefl

The FwFunc struct contains an function with a char pointer argument, for which a *const u8 pointer was used. This is not really the "propper" type for this, so use a *const kernel::ffi::c_char pointer instad.

This has no real functionality changes, since kernel::ffi::c_char is a type alias to u8 anyways.

This used to cause problems on 6.13 when building for 32 bit arm (with my patches), since rust mapped c_char to i8 instead.

Now that I read this again:

Isn't it the other way around? For arm32 c_char was mapped to u8, but FwFunc expected i8 (since that's what c_char was mapped to for all other architectures that are supported in v6.13).

Can you please clarify this in the commit message?

#7 lore raw
Quoted 12 lines · Danilo Krummrich
Quoted 10 lines · Christian Schrefl

The FwFunc struct contains an function with a char pointer argument, for which a *const u8 pointer was used. This is not really the "propper" type for this, so use a *const kernel::ffi::c_char pointer instad.

This has no real functionality changes, since kernel::ffi::c_char is a type alias to u8 anyways.

This used to cause problems on 6.13 when building for 32 bit arm (with my patches), since rust mapped c_char to i8 instead.

Quoted 7 lines

Now that I read this again:

Isn't it the other way around? For arm32 c_char was mapped to u8, but FwFunc expected i8 (since that's what c_char was mapped to for all other architectures that are supported in v6.13).

Can you please clarify this in the commit message?

Ah sorry I got confused because by the change in 1bae8729e50a ("rust:
map long to isize and char to u8").

How about changing that section to:
....
This has no real functionality changes, since now kernel::ffi::c_char is now a type alias to u8 anyways, but before commit 1bae8729e50a ("rust:
map long to isize and char to u8") the concrete type of core::ffi::c_char depended on the architecture (However all supported architectures at the time mapped to i8).

This causes problems on v6.13 when building for 32 bit arm (with my patches), since back then *const u8 was used in the function argument and the function that bindgen generated used *const core::ffi::c_char which Rust mapped to u8 on 32 bit arm.

This caused the following build error:
...

Cheers
Christian

maintainer
#8 lore raw
Quoted 32 lines · Christian Schrefl
Quoted 12 lines · Danilo Krummrich
Quoted 10 lines · Christian Schrefl

The FwFunc struct contains an function with a char pointer argument, for which a *const u8 pointer was used. This is not really the "propper" type for this, so use a *const kernel::ffi::c_char pointer instad.

This has no real functionality changes, since kernel::ffi::c_char is a type alias to u8 anyways.

This used to cause problems on 6.13 when building for 32 bit arm (with my patches), since rust mapped c_char to i8 instead.

Quoted 7 lines

Now that I read this again:

Isn't it the other way around? For arm32 c_char was mapped to u8, but FwFunc expected i8 (since that's what c_char was mapped to for all other architectures that are supported in v6.13).

Can you please clarify this in the commit message?

Ah sorry I got confused because by the change in 1bae8729e50a ("rust:
map long to isize and char to u8").

How about changing that section to:
....
This has no real functionality changes, since now kernel::ffi::c_char is now a type alias to u8 anyways, but before commit 1bae8729e50a ("rust:
map long to isize and char to u8") the concrete type of core::ffi::c_char depended on the architecture (However all supported architectures at the time mapped to i8).

Quoted 2 lines

This causes problems on v6.13 when building for 32 bit arm (with my patches), since back then *const u8 was used in the function argument

I'd put slightly more emphasis on the fact that there's no problem in v6.13 without addition your additional patches to introduce 32 bit arm support.

Otherwise, sounds good!

Thanks,
Danilo

#9 lore raw
In reply to Christian Schrefl
Quoted 3 lines · Christian Schrefl

This causes problems on v6.13 when building for 32 bit arm (with my patches), since back then *const u8 was used in the function argument

In v6.13, the function argument used i8 -- if it were u8, you wouldn't have had an issue, no?

Nowadays, in 6.13.y, everything is u8, so you shouldn't have a build error with your patches.

Cheers,
Miguel

#10 lore raw
Quoted 8 lines · Miguel Ojeda
Quoted 3 lines · Christian Schrefl

This causes problems on v6.13 when building for 32 bit arm (with my patches), since back then *const u8 was used in the function argument

In v6.13, the function argument used i8 -- if it were u8, you wouldn't have had an issue, no?

Yes I've sent a new version of this section in the thread with Danilo.

I'll send it out as V3 soon.

Quoted 6 lines

Nowadays, in 6.13.y, everything is u8, so you shouldn't have a build error with your patches.

Cheers,
Miguel

#12 lore raw
In reply to Miguel Ojeda
Quoted 9 lines · Miguel Ojeda
Quoted 3 lines · Christian Schrefl

This causes problems on v6.13 when building for 32 bit arm (with my patches), since back then *const u8 was used in the function argument

In v6.13, the function argument used i8 -- if it were u8, you wouldn't have had an issue, no?

Ah right thats a typo. Should have said i8`.

Quoted 3 lines

Nowadays, in 6.13.y, everything is u8, so you shouldn't have a build error with your patches.

I'll mention it explicitly that its not a problem in stable 6.13, only in the v6.13 tag.

Sorry for the confusion, Thunderbird showed this as a response to the original patch mail, so I thought you were referring to that.

Cheers
Christian

#13 lore raw
Quoted 3 lines · Christian Schrefl

Sorry for the confusion, Thunderbird showed this as a response to the original patch mail, so I thought you were referring to that.

No worries! Lore shows it as replying to the right message, so I assume it is Thunderbird-related, yeah.

Thanks!

Cheers,
Miguel