Skip to contentSkip to conversation

rust: bug: fix `dead_code` warning in `warn_on!`

PATCH In review Miguel Ojeda 4 participants 9 messages Aug 7 – Aug 9

Subsystems (4)
lore
Share & tools
lore

Miguel Ojeda

author maintainer
#0 lore raw

PATCH 1/1 rust: bug: fix `dead_code` warning in `warn_on!`

On arm and loongarch64, or under CONFIG_BUG=n, the upcoming SRCU abstractions from commit

59cf3a5bda61 ("rust: sync: add SRCU abstraction")

use warn_on!, which in turn warns:

    error: constant `WARN_ON_FLAGS` is never used
       --> rust/kernel/bug.rs:126:19
        |
    126 |               const WARN_ON_FLAGS: u32 = $crate::bug::bugflag_taint($crate::bindings::TAINT_WARN);
        |                     ^^^^^^^^^^^^^
        |
       ::: rust/kernel/sync/srcu.rs:106:12
        |
    106 |           if crate::warn_on!(
        |  ____________-
    107 | |             // SAFETY: By the type invariants, `self` contains a valid and pinned `struct srcu_struct`
    108 | |             // and `srcu_readers_active()` only checks the active reader count.
    109 | |             unsafe { bindings::srcu_readers_active(ptr) }
    110 | |         ) {
        | |_________- in this macro invocation
        |
        = note: `-D dead-code` implied by `-D warnings`
        = help: to override `-D warnings` add `#[allow(dead_code)]`
        = note: this error originates in the macro `crate::warn_on` (in Nightly builds, run with -Z macro-backtrace for more info)

The reason is that warn_flags! discards both arguments in those cases, and so the const becomes unused. And since it is a macro, it went unnoticed so far.

Thus fix it by simplifying by removing the const.

Fixes dff64b072708 ("rust: Add warn_on macro")

---
Either this or adding an underscore.

 rust/kernel/bug.rs | 7 ++++---
 1 file changed, 4 insertions(+), 3 deletions(-)
rust/kernel/bug.rs +4 -3
index ed943960f851..874adb93cdfe 100644
@@ -123,9 +123,10 @@ macro_rules! warn_on {
         const _COND_STR: &str = file!();
 
         if cond {
-            const WARN_ON_FLAGS: u32 = $crate::bug::bugflag_taint($crate::bindings::TAINT_WARN);
-
-            $crate::warn_flags!(_COND_STR, WARN_ON_FLAGS);
+            $crate::warn_flags!(
+                _COND_STR,
+                $crate::bug::bugflag_taint($crate::bindings::TAINT_WARN)
+            );
         }
         cond
     }};

base-commit: dc01dfb37b34beeefcfe1c3055364d41a4070c7e --
2.55.0

#1 lore raw
Quoted 32 lines · Miguel Ojeda

On arm and loongarch64, or under CONFIG_BUG=n, the upcoming SRCU abstractions from commit

59cf3a5bda61 ("rust: sync: add SRCU abstraction")

use warn_on!, which in turn warns:

    error: constant `WARN_ON_FLAGS` is never used
       --> rust/kernel/bug.rs:126:19
        |
    126 |               const WARN_ON_FLAGS: u32 = $crate::bug::bugflag_taint($crate::bindings::TAINT_WARN);
        |                     ^^^^^^^^^^^^^
        |
       ::: rust/kernel/sync/srcu.rs:106:12
        |
    106 |           if crate::warn_on!(
        |  ____________-
    107 | |             // SAFETY: By the type invariants, `self` contains a valid and pinned `struct srcu_struct`
    108 | |             // and `srcu_readers_active()` only checks the active reader count.
    109 | |             unsafe { bindings::srcu_readers_active(ptr) }
    110 | |         ) {
        | |_________- in this macro invocation
        |
        = note: `-D dead-code` implied by `-D warnings`
        = help: to override `-D warnings` add `#[allow(dead_code)]`
        = note: this error originates in the macro `crate::warn_on` (in Nightly builds, run with -Z macro-backtrace for more info)

The reason is that warn_flags! discards both arguments in those cases, and so the const becomes unused. And since it is a macro, it went unnoticed so far.

Quoted 5 lines

Thus fix it by simplifying by removing the const.

Fixes dff64b072708 ("rust: Add warn_on macro")

Sorry, I forgot the Cc to Tomonori:

Cc FUJITA Tomonori <[email protected]>

Cheers,
Miguel

#2 lore raw
In reply to Miguel Ojeda
Quoted 61 lines · On Sat, 8 Aug 2026 00:14:29 +0200 Miguel Ojeda <[email protected]> wrote:

On arm and loongarch64, or under CONFIG_BUG=n, the upcoming SRCU abstractions from commit

59cf3a5bda61 ("rust: sync: add SRCU abstraction")

use warn_on!, which in turn warns:

    error: constant `WARN_ON_FLAGS` is never used
       --> rust/kernel/bug.rs:126:19
        |
    126 |               const WARN_ON_FLAGS: u32 = $crate::bug::bugflag_taint($crate::bindings::TAINT_WARN);
        |                     ^^^^^^^^^^^^^
        |
       ::: rust/kernel/sync/srcu.rs:106:12
        |
    106 |           if crate::warn_on!(
        |  ____________-
    107 | |             // SAFETY: By the type invariants, `self` contains a valid and pinned `struct srcu_struct`
    108 | |             // and `srcu_readers_active()` only checks the active reader count.
    109 | |             unsafe { bindings::srcu_readers_active(ptr) }
    110 | |         ) {
        | |_________- in this macro invocation
        |
        = note: `-D dead-code` implied by `-D warnings`
        = help: to override `-D warnings` add `#[allow(dead_code)]`
        = note: this error originates in the macro `crate::warn_on` (in Nightly builds, run with -Z macro-backtrace for more info)

The reason is that warn_flags! discards both arguments in those cases, and so the const becomes unused. And since it is a macro, it went unnoticed so far.

Thus fix it by simplifying by removing the const.

Fixes dff64b072708 ("rust: Add warn_on macro")

---
Either this or adding an underscore.

 rust/kernel/bug.rs | 7 ++++---
 1 file changed, 4 insertions(+), 3 deletions(-)
rust/kernel/bug.rs +4 -3
index ed943960f851..874adb93cdfe 100644
@@ -123,9 +123,10 @@ macro_rules! warn_on {
         const _COND_STR: &str = file!();
 
         if cond {
-            const WARN_ON_FLAGS: u32 = $crate::bug::bugflag_taint($crate::bindings::TAINT_WARN);
-
-            $crate::warn_flags!(_COND_STR, WARN_ON_FLAGS);
+            $crate::warn_flags!(
+                _COND_STR,
+                $crate::bug::bugflag_taint($crate::bindings::TAINT_WARN)
+            );
         }
         cond
     }};
Quoted 3 lines

base-commit: dc01dfb37b34beeefcfe1c3055364d41a4070c7e --
2.55.0

I sent a fix for this bug a few days ago:

https://lore.kernel.org/all/[email protected]/

If you prefer the simpler version, v1 might work.

https://lore.kernel.org/all/[email protected]/

#4 lore raw
Quoted 9 lines · On Sat, 8 Aug 2026 01:21:23 +0200 Miguel Ojeda…
Quoted 4 lines · FUJITA Tomonori

I sent a fix for this bug a few days ago:

https://lore.kernel.org/all/[email protected]/

Ah, right, sorry! I also saw another one with rusttest for arm64 and riscv64. Did you see that one? We probably need a cfg(testlib) split.

Is this the one you saw?

error: invalid instruction mnemonic 'brk'
--> rust/kernel/bug.rs:63:17

   |
63 | /                 concat!(
64 | |                     "/* {size} */",
65 | |                     include!(concat!(env!("OBJTREE"), "/rust/kernel/generated_arch_warn_asm.rs")),
66 | |                     include!(concat!(env!("OBJTREE"), "/rust/kernel/generated_arch_reachable_asm.rs")));
   | |_______________________________________________________________________________________________________^
   |
note: instantiated into assembly here
  --> <inline asm>:1:115
   |
 1 |     /* 8 */.pushsection __bug_table,"aw"; .align 2; 14470: .long 14471f - .;.short 2305;.align 2; .popsection; 14471:brk 0x800
   |                                                                                                                      ^^^

It shows up when make rusttest runs with ARCH= set to an architecture other than the host's.

I posted a cfg(testlib) split for it:

https://lore.kernel.org/all/[email protected]/

While at it, a question: what is rusttest meant to cover when ARCH= is not the host's?

#5 lore raw
Quoted 2 lines · FUJITA Tomonori

Is this the one you saw?

Yeah, exactly.

Quoted 3 lines

I posted a cfg(testlib) split for it:

https://lore.kernel.org/all/[email protected]/

Thanks! It seems good to me.

Quoted 2 lines

While at it, a question: what is rusttest meant to cover when ARCH= is not the host's?

rusttest is meant to run tests on the host, regardless of the target.

There aren't many tests there nowadays, since most of them moved to the actual kernel when we got support for that, but we can still do there things like testing proc macros.

We could also use it for more things in the future, perhaps...

Cheers,
Miguel

#6 lore raw
Quoted 5 lines · On Sat, 8 Aug 2026 21:40:49 +0200 Miguel Ojeda…
Quoted 3 lines

I posted a cfg(testlib) split for it:

https://lore.kernel.org/all/[email protected]/

Thanks! It seems good to me.

Thanks!

Quoted 4 lines
Quoted 2 lines

While at it, a question: what is rusttest meant to cover when ARCH= is not the host's?

rusttest is meant to run tests on the host, regardless of the target.

Then jump_label.rs needs the same treatment?

arch_static_branch! embeds the generated arch asm too. It is not a problem today only because, like warn_on! until now, it has no user inside the kernel crate itself.

#7 lore raw
Quoted 6 lines · FUJITA Tomonori

Then jump_label.rs needs the same treatment?

arch_static_branch! embeds the generated arch asm too. It is not a problem today only because, like warn_on! until now, it has no user inside the kernel crate itself.

Yeah, I think we should do it to avoid surprises later.

Good eyes!

Cheers,
Miguel

reviewer
#8 lore raw
In reply to Miguel Ojeda
Quoted 9 lines · Miguel Ojeda
Quoted 4 lines · FUJITA Tomonori

I sent a fix for this bug a few days ago:

https://lore.kernel.org/all/[email protected]/

Ah, right, sorry! I also saw another one with rusttest for arm64 and riscv64. Did you see that one? We probably need a cfg(testlib) split.

I am not sure how much value does rusttest still provide, given that KUnit can test much more stuff in an actual envirionment. The fact that we have to do special casing does suggest to me that we might want to remove it.

Best,
Gary