[PATCH] rust_binder: treat ESRCH as BR_DEAD_REPLY instead of BR_FAILED_REPLY

From: Alice Ryhl

Date: Wed Nov 18 2026 - 06:24:48 EST


When sending a message concurrently with it exiting, the transaction may
fail with ESRCH if the timing is right. Since this is delivered to
userspace as BR_FAILED_REPLY, userspace thinks that something went wrong
with the transaction, but actually the target process just died. This
usually leads to the sender aborting.

Cc: stable@xxxxxxxxxxxxxxx
Fixes: eafedbc7c050 ("rust_binder: add Rust Binder driver")
Signed-off-by: Alice Ryhl <aliceryhl@xxxxxxxxxx>
---
drivers/android/binder/error.rs | 6 +++++-
1 file changed, 5 insertions(+), 1 deletion(-)

diff --git a/drivers/android/binder/error.rs b/drivers/android/binder/error.rs
index acf0aa09c555..00f1519a8d77 100644
--- a/drivers/android/binder/error.rs
+++ b/drivers/android/binder/error.rs
@@ -78,8 +78,12 @@ pub(crate) fn new_frozen_oneway() -> Self {
impl From<Error> for BinderError {
#[track_caller]
fn from(source: Error) -> Self {
+ let reply = match source {
+ ESRCH => BR_DEAD_REPLY,
+ _ => BR_FAILED_REPLY,
+ };
Self {
- reply: BR_FAILED_REPLY,
+ reply,
source: Some(source),
line: ErrorLocation::caller(),
}

---
base-commit: 20eef29b7c8465ac98d7dd6d6f2ac07a8d8056ca
change-id: 20261029-binder-esrch-is-dead-8109ceb9115c

Best regards,
--
Alice Ryhl <aliceryhl@xxxxxxxxxx>