From 6caae86b7c7591b1e2902a9002b7649fe99022e5 Mon Sep 17 00:00:00 2001
From: xie xingguo <xie.xingguo@zte.com.cn>
Date: Fri, 15 Sep 2017 22:28:40 +0800
Subject: [PATCH] os/bluestore: propagate read-EIO for aio

Signed-off-by: xie xingguo <xie.xingguo@zte.com.cn>
(cherry picked from commit 301912603789d78e3560da84df1e337edd046e46)
---
 src/os/bluestore/BlockDevice.h   | 14 ++++++++++++--
 src/os/bluestore/BlueStore.cc    |  7 ++++++-
 src/os/bluestore/KernelDevice.cc | 32 +++++++++++++++++---------------
 3 files changed, 35 insertions(+), 18 deletions(-)

diff --git a/src/os/bluestore/BlockDevice.h b/src/os/bluestore/BlockDevice.h
index 5d511ddbb5708..280e650cb3653 100644
--- a/src/os/bluestore/BlockDevice.h
+++ b/src/os/bluestore/BlockDevice.h
@@ -32,6 +32,7 @@ struct IOContext {
 private:
   std::mutex lock;
   std::condition_variable cond;
+  int r = 0;
 
 public:
   CephContext* cct;
@@ -46,9 +47,10 @@ public:
   std::list<aio_t> running_aios;    ///< submitting or submitted
   std::atomic_int num_pending = {0};
   std::atomic_int num_running = {0};
+  bool allow_eio;
 
-  explicit IOContext(CephContext* cct, void *p)
-    : cct(cct), priv(p)
+  explicit IOContext(CephContext* cct, void *p, bool allow_eio = false)
+    : cct(cct), priv(p), allow_eio(allow_eio)
     {}
 
   // no copying
@@ -76,6 +78,14 @@ public:
       --num_running;
     }
   }
+
+  void set_return_value(int _r) {
+    r = _r;
+  }
+
+  int get_return_value() const {
+    return r;
+  }
 };
 
 
diff --git a/src/os/bluestore/BlueStore.cc b/src/os/bluestore/BlueStore.cc
index 2a414c1efc0d8..37ec919daba7a 100644
--- a/src/os/bluestore/BlueStore.cc
+++ b/src/os/bluestore/BlueStore.cc
@@ -6614,7 +6614,7 @@ int BlueStore::_do_read(
                                     // measure the whole block below.
                                     // The error isn't that much...
   vector<bufferlist> compressed_blob_bls;
-  IOContext ioc(cct, NULL);
+  IOContext ioc(cct, NULL, true); // allow EIO
   for (auto& p : blobs2read) {
     BlobRef bptr = p.first;
     dout(20) << __func__ << "  blob " << *bptr << std::hex
@@ -6703,6 +6703,11 @@ int BlueStore::_do_read(
     bdev->aio_submit(&ioc);
     dout(20) << __func__ << " waiting for aio" << dendl;
     ioc.aio_wait();
+    r = ioc.get_return_value();
+    if (r < 0) {
+      assert(r == -EIO); // no other errors allowed
+      return -EIO;
+    }
   }
   logger->tinc(l_bluestore_read_wait_aio_lat, ceph_clock_now() - start);
 
diff --git a/src/os/bluestore/KernelDevice.cc b/src/os/bluestore/KernelDevice.cc
index 4dd2f9ca03e47..b3cc7795d9a92 100644
--- a/src/os/bluestore/KernelDevice.cc
+++ b/src/os/bluestore/KernelDevice.cc
@@ -371,21 +371,23 @@ void KernelDevice::_aio_thread()
 	io_since_flush.store(true);
 
 	int r = aio[i]->get_return_value();
-       if (r < 0) {
-         derr << __func__ << " got " << cpp_strerror(r) << dendl;
-         assert(0 == "got unexpected error from io_getevents");
-       }
-	
-       if (aio[i]->length != (uint64_t)r) {
-         derr << "aio to " << aio[i]->offset << "~" << aio[i]->length
-              << " but returned: " << r << dendl;
-         assert(0 == "unexpected aio error");
-       }
-
-       dout(10) << __func__ << " finished aio " << aio[i] << " r " << r
-                << " ioc " << ioc
-                << " with " << (ioc->num_running.load() - 1)
-                << " aios left" << dendl;
+        if (r < 0) {
+          derr << __func__ << " got " << cpp_strerror(r) << dendl;
+          if (ioc->allow_eio && r == -EIO) {
+            ioc->set_return_value(r);
+          } else {
+            assert(0 == "got unexpected error from io_getevents");
+          }
+        } else if (aio[i]->length != (uint64_t)r) {
+          derr << "aio to " << aio[i]->offset << "~" << aio[i]->length
+               << " but returned: " << r << dendl;
+          assert(0 == "unexpected aio error");
+        }
+
+        dout(10) << __func__ << " finished aio " << aio[i] << " r " << r
+                 << " ioc " << ioc
+                 << " with " << (ioc->num_running.load() - 1)
+                 << " aios left" << dendl;
 
 	// NOTE: once num_running and we either call the callback or
 	// call aio_wake we cannot touch ioc or aio[] as the caller
-- 
2.39.5