[PATCH v2] rust_binder: avoid name mangling for get_work[_local]

Alice Ryhl posted 1 patch 1 month, 1 week ago
drivers/android/binder/process.rs | 3 +++
drivers/android/binder/thread.rs  | 6 ++++++
2 files changed, 9 insertions(+)
[PATCH v2] rust_binder: avoid name mangling for get_work[_local]
Posted by Alice Ryhl 1 month, 1 week ago
Currently ps -A shows processes waiting on schedule() in functions with
names such as do_epoll_wait, wait_woken, and the impeccably named
_RNvMs2_NtCs8QPsHWIn21X_16rust_binder_main6threadNtB5_6Thread8get_work.

To improve how ps output looks, give explicit non-mangled names to the
functions where Rust Binder calls schedule(), since these are the most
likely places to show up on ps output.

The name of rust_binder_waitlcl is truncated instead of using _local
suffix because rust_binder_wait_local is sufficiently long that ps shows
unaligned output.

This is intended to be a temporary workaround until we find a better
solution. Adding #[export_name] to every Rust function that calls
schedule() is not a great long-term solution.

Suggested-by: Matthew Maurer <mmaurer@google.com>
Signed-off-by: Alice Ryhl <aliceryhl@google.com>
---
Changes in v2:
- Rebase.
- Clarify that this is a workaround.
- Link to v1: https://lore.kernel.org/r/20260213-rust-binder-ps-v1-1-6b90f83c0b4d@google.com
---
 drivers/android/binder/process.rs | 3 +++
 drivers/android/binder/thread.rs  | 6 ++++++
 2 files changed, 9 insertions(+)

diff --git a/drivers/android/binder/process.rs b/drivers/android/binder/process.rs
index 41de5593197cd29b71d1cb49829a1ff878bab249..f62f626f928e7ea4965d03c92cc988f344c6ae81 100644
--- a/drivers/android/binder/process.rs
+++ b/drivers/android/binder/process.rs
@@ -1442,6 +1442,9 @@ pub(crate) fn drop_outstanding_txn(&self) {
         }
     }
 
+    // #[export_name] is a temporary workaround so that ps output does not become unreadable from
+    // mangled symbol names.
+    #[export_name = "rust_binder_freeze"]
     pub(crate) fn ioctl_freeze(&self, info: &BinderFreezeInfo) -> Result {
         if info.enable == 0 {
             let msgs = self.prepare_freeze_messages()?;
diff --git a/drivers/android/binder/thread.rs b/drivers/android/binder/thread.rs
index 0b62d24b2118abe10577029a0129887e4ee8c6ba..6f197be0fa75d9d485c6cd659fa28c872a88af93 100644
--- a/drivers/android/binder/thread.rs
+++ b/drivers/android/binder/thread.rs
@@ -513,6 +513,9 @@ pub(crate) fn has_current_transaction(&self) -> bool {
     /// Attempts to fetch a work item from the thread-local queue. The behaviour if the queue is
     /// empty depends on `wait`: if it is true, the function waits for some work to be queued (or a
     /// signal); otherwise it returns indicating that none is available.
+    // #[export_name] is a temporary workaround so that ps output does not become unreadable from
+    // mangled symbol names.
+    #[export_name = "rust_binder_waitlcl"]
     fn get_work_local(self: &Arc<Self>, wait: bool) -> Result<Option<DLArc<dyn DeliverToRead>>> {
         {
             let mut inner = self.inner.lock();
@@ -551,6 +554,9 @@ fn get_work_local(self: &Arc<Self>, wait: bool) -> Result<Option<DLArc<dyn Deliv
     ///
     /// This must only be called when the thread is not participating in a transaction chain. If it
     /// is, the local version (`get_work_local`) should be used instead.
+    // #[export_name] is a temporary workaround so that ps output does not become unreadable from
+    // mangled symbol names.
+    #[export_name = "rust_binder_wait"]
     fn get_work(self: &Arc<Self>, wait: bool) -> Result<Option<DLArc<dyn DeliverToRead>>> {
         // Try to get work from the thread's work queue, using only a local lock.
         {

---
base-commit: 2961f841b025fb234860bac26dfb7fa7cb0fb122
change-id: 20260213-rust-binder-ps-46d9d060243a

Best regards,
-- 
Alice Ryhl <aliceryhl@google.com>
Re: [PATCH v2] rust_binder: avoid name mangling for get_work[_local]
Posted by Gary Guo 1 month, 1 week ago
On 2026-02-19 11:27, Alice Ryhl wrote:
> Currently ps -A shows processes waiting on schedule() in functions with
> names such as do_epoll_wait, wait_woken, and the impeccably named
> _RNvMs2_NtCs8QPsHWIn21X_16rust_binder_main6threadNtB5_6Thread8get_work.
> 
> To improve how ps output looks, give explicit non-mangled names to the
> functions where Rust Binder calls schedule(), since these are the most
> likely places to show up on ps output.
> 
> The name of rust_binder_waitlcl is truncated instead of using _local
> suffix because rust_binder_wait_local is sufficiently long that ps 
> shows
> unaligned output.
> 
> This is intended to be a temporary workaround until we find a better
> solution. Adding #[export_name] to every Rust function that calls
> schedule() is not a great long-term solution.
> 
> Suggested-by: Matthew Maurer <mmaurer@google.com>
> Signed-off-by: Alice Ryhl <aliceryhl@google.com>

Acked-by: Gary Guo <gary@garyguo.net>

> ---
> Changes in v2:
> - Rebase.
> - Clarify that this is a workaround.
> - Link to v1: 
> https://lore.kernel.org/r/20260213-rust-binder-ps-v1-1-6b90f83c0b4d@google.com
> ---
>  drivers/android/binder/process.rs | 3 +++
>  drivers/android/binder/thread.rs  | 6 ++++++
>  2 files changed, 9 insertions(+)
> 
> diff --git a/drivers/android/binder/process.rs 
> b/drivers/android/binder/process.rs
> index 
> 41de5593197cd29b71d1cb49829a1ff878bab249..f62f626f928e7ea4965d03c92cc988f344c6ae81 
> 100644
> --- a/drivers/android/binder/process.rs
> +++ b/drivers/android/binder/process.rs
> @@ -1442,6 +1442,9 @@ pub(crate) fn drop_outstanding_txn(&self) {
>          }
>      }
> 
> +    // #[export_name] is a temporary workaround so that ps output does 
> not become unreadable from
> +    // mangled symbol names.
> +    #[export_name = "rust_binder_freeze"]
>      pub(crate) fn ioctl_freeze(&self, info: &BinderFreezeInfo) -> 
> Result {
>          if info.enable == 0 {
>              let msgs = self.prepare_freeze_messages()?;
> diff --git a/drivers/android/binder/thread.rs 
> b/drivers/android/binder/thread.rs
> index 
> 0b62d24b2118abe10577029a0129887e4ee8c6ba..6f197be0fa75d9d485c6cd659fa28c872a88af93 
> 100644
> --- a/drivers/android/binder/thread.rs
> +++ b/drivers/android/binder/thread.rs
> @@ -513,6 +513,9 @@ pub(crate) fn has_current_transaction(&self) -> 
> bool {
>      /// Attempts to fetch a work item from the thread-local queue. The 
> behaviour if the queue is
>      /// empty depends on `wait`: if it is true, the function waits for 
> some work to be queued (or a
>      /// signal); otherwise it returns indicating that none is 
> available.
> +    // #[export_name] is a temporary workaround so that ps output does 
> not become unreadable from
> +    // mangled symbol names.
> +    #[export_name = "rust_binder_waitlcl"]
>      fn get_work_local(self: &Arc<Self>, wait: bool) -> 
> Result<Option<DLArc<dyn DeliverToRead>>> {
>          {
>              let mut inner = self.inner.lock();
> @@ -551,6 +554,9 @@ fn get_work_local(self: &Arc<Self>, wait: bool) -> 
> Result<Option<DLArc<dyn Deliv
>      ///
>      /// This must only be called when the thread is not participating 
> in a transaction chain. If it
>      /// is, the local version (`get_work_local`) should be used 
> instead.
> +    // #[export_name] is a temporary workaround so that ps output does 
> not become unreadable from
> +    // mangled symbol names.
> +    #[export_name = "rust_binder_wait"]
>      fn get_work(self: &Arc<Self>, wait: bool) -> 
> Result<Option<DLArc<dyn DeliverToRead>>> {
>          // Try to get work from the thread's work queue, using only a 
> local lock.
>          {
> 
> ---
> base-commit: 2961f841b025fb234860bac26dfb7fa7cb0fb122
> change-id: 20260213-rust-binder-ps-46d9d060243a
> 
> Best regards,