From c8822ddea3d311285f766ba2957778ef5ac51b04 Mon Sep 17 00:00:00 2001
From: Matthew Jackson <1085847+MattJackson@users.noreply.github.com>
Date: Tue, 30 Jun 2026 07:19:03 -0700
Subject: [PATCH] error: treat IoError/DeviceNotFound as SCSI transport
 failures (dead bus)
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

is_scsi_transport_failure recognized only the 0xFF SCSI sentinel, so the two
NON-SCSI dead-bus faults the Linux SG_IO transport returns — Error::IoError
(ioctl(SG_IO) == -1: ENODEV/EIO on an unplugged bridge) and Error::DeviceNotFound
(fd gone) — were classified as recoverable bad sectors. Every caller
(sweep read_error, patch, single-pass fill_extents) would then zero-fill /
skip-ahead across a wedged bus instead of aborting the pass and re-enumerating
the bridge — potentially zero-filling a huge span. Add both to the classifier so
a dead bus aborts the pass everywhere. (Companion to the unlock_bridge fix, which
maps the same two faults to 0xFF across the freemkv-unlock seam.)
---
 src/error.rs | 36 +++++++++++++++++++++++++++++++-----
 1 file changed, 31 insertions(+), 5 deletions(-)

diff --git a/src/error.rs b/src/error.rs
index a206a5b..6425be5 100644
--- a/src/error.rs
+++ b/src/error.rs
@@ -879,6 +879,15 @@ impl Error {
                 status: Some(crate::scsi::SCSI_STATUS_TRANSPORT_FAILURE),
                 ..
             }
+        ) || matches!(
+            // A failed `ioctl(SG_IO)` (Error::IoError, e.g. ENODEV/EIO on an
+            // unplugged USB bridge) and a vanished device (Error::DeviceNotFound,
+            // fd gone) are dead-bus / transport-layer faults too — NOT recoverable
+            // bad sectors. Treat them as transport failures so sweep / patch /
+            // fill_extents abort the pass and re-enumerate the bridge instead of
+            // zero-filling every read against a wedged device.
+            self,
+            Error::IoError { .. } | Error::DeviceNotFound { .. }
         )
     }
 
@@ -1375,11 +1384,13 @@ mod tests {
         }
     }
 
-    /// is_scsi_transport_failure returns true only for SCSI_STATUS_TRANSPORT_FAILURE (0xFF).
-    /// Spec: comment on SCSI_STATUS_TRANSPORT_FAILURE says "synthesised sentinel: the
-    ///       transport never delivered a SCSI status byte".
+    /// is_scsi_transport_failure is true for the 0xFF SCSI sentinel AND for the
+    /// non-SCSI dead-bus faults (Error::IoError from a failed ioctl(SG_IO),
+    /// Error::DeviceNotFound from a vanished fd) — but NEVER for a real SCSI
+    /// reply (CHECK CONDITION) or unrelated errors.
     /// Mutation: testing against 0x02 (CHECK CONDITION) would wrongly mark CHECK
-    ///           CONDITION replies as transport failures.
+    ///           CONDITION replies as transport failures; dropping the IoError/
+    ///           DeviceNotFound arm would let a dead bus zero-fill the disc.
     #[test]
     fn is_scsi_transport_failure_only_for_0xff() {
         use crate::scsi::SCSI_STATUS_TRANSPORT_FAILURE;
@@ -1403,7 +1414,22 @@ mod tests {
         };
         assert!(!cc.is_scsi_transport_failure());
 
-        // False for non-SCSI errors.
+        // True: non-SCSI dead-bus faults — a failed ioctl(SG_IO) and a vanished
+        // device are transport-layer failures, not recoverable bad sectors.
+        assert!(
+            Error::IoError {
+                source: std::io::Error::from(std::io::ErrorKind::NotConnected)
+            }
+            .is_scsi_transport_failure()
+        );
+        assert!(
+            Error::DeviceNotFound {
+                path: "/dev/sg9".into()
+            }
+            .is_scsi_transport_failure()
+        );
+
+        // False for unrelated errors.
         assert!(!Error::Halted.is_scsi_transport_failure());
     }