feat(nvme): On command completion handle zoned nvme errors

Zoned related CommandSpecificStatusCodes's should not take out the
nexus.

Signed-off-by: Dennis Maisenbacher <[email protected]>

io-engine/src/bdev/nexus/nexus_io.rs

@@ -33,6 +33,7 @@ use crate::core::{
     DeviceCommand,
     GenericStatusCode,
     IoCompletionStatus,
+    is_zoned_nvme_error,
     IoStatus,
     IoSubmissionFailure,
     IoType,
@@ -193,7 +194,6 @@ impl<'n> NexusBio<'n> {
         if success {
             self.ok_checked();
         } else {
-            // IO failure, mark the IO failed and take the child out
             error!(
                 "{:?}: IO completion for '{}' failed: {:?}, ctx={:?}",
                 self,
@@ -202,8 +202,15 @@ impl<'n> NexusBio<'n> {
                 self.ctx()
             );
             self.ctx_mut().status = IoStatus::Failed;
-            self.ctx_mut().must_fail = true;
-            self.handle_failure(child, status);
+            if is_zoned_nvme_error(status) {
+                // Don't take zoned child out on zoned related nvme errors
+                self.ctx_mut().must_fail = false;
+                self.ok_checked();
+            } else {
+                // IO failure, mark the IO failed and take the child out
+                self.ctx_mut().must_fail = true;
+                self.handle_failure(child, status);
+            }
         }
     }
 

io-engine/src/core/mod.rs

@@ -59,7 +59,7 @@ pub use lock::{
 };
 pub use runtime::spawn;
 pub use share::{Protocol, PtplProps, Share, ShareProps, UpdateProps};
-pub use spdk_rs::{cpu_cores, GenericStatusCode, IoStatus, IoType, NvmeStatus};
+pub use spdk_rs::{cpu_cores, GenericStatusCode, IoStatus, IoType, NvmeStatus, CommandSpecificStatusCode};
 pub use thread::Mthread;
 
 use crate::subsys::NvmfError;
@@ -301,6 +301,23 @@ impl From<NvmeStatus> for IoCompletionStatus {
     }
 }
 
+pub fn is_zoned_nvme_error(status: IoCompletionStatus) -> bool {
+    match status {
+        IoCompletionStatus::NvmeError(NvmeStatus::CommandSpecific(cssc)) => match cssc {
+            CommandSpecificStatusCode::ZonedBoundaryError => true,
+            CommandSpecificStatusCode::ZoneIsFull => true,
+            CommandSpecificStatusCode::ZoneIsReadOnly => true,
+            CommandSpecificStatusCode::ZoneIsOffline => true,
+            CommandSpecificStatusCode::ZoneInvalidWrite => true,
+            CommandSpecificStatusCode::TooManyActiveZones => true,
+            CommandSpecificStatusCode::TooManyOpenZones => true,
+            CommandSpecificStatusCode::InvalidZoneStateTransition => true,
+            _ => false,
+        },
+        _ => false,
+    }
+}
+
 // TODO move this elsewhere ASAP
 pub static PAUSING: AtomicUsize = AtomicUsize::new(0);
 pub static PAUSED: AtomicUsize = AtomicUsize::new(0);