Fix for the Linux Kernel v6.2 (Fedora 37/38) (#258)

* Enable Fedora 38 * Added compat for bio_set_op_attrs * Fixed kernel panic on dereferencing the wrong address * Increased volume sizes for the new mkfs.xfs (needs at least 300 MB per partition) * Increased timeouts in CI tests Closes #247
elastio · May 10, 2023 · bde38ae · bde38ae
1 parent 2ecc357
commit bde38ae
Show file tree

Hide file tree

Showing 9 changed files with 51 additions and 28 deletions.
diff --git a/.github/buildbox/Vagrantfile b/.github/buildbox/Vagrantfile
@@ -38,14 +38,14 @@ DISTRO['amd64'] = [ "debian8", "debian9", "debian10", "debian11",
                     "amazon2", "amazon2023",
                     "centos7", "centos8", "centos9",
                     "alma8", "alma9",
-                    "fedora31", "fedora32", "fedora34", "fedora35", "fedora36", "fedora37",
+                    "fedora31", "fedora32", "fedora34", "fedora35", "fedora36", "fedora37", "fedora38",
                     "ubuntu2004", "ubuntu2204" ]
 
 DISTRO['arm64'] = [ "debian10", "debian11",
                     "amazon2", "amazon2023",
                     "centos8", "centos9",
                     "alma8", "alma9",
-                    "fedora35", "fedora36", "fedora37",
+                    "fedora35", "fedora36", "fedora37", "fedora38",
                     "ubuntu2204" ]
 
 # NOTE: This environment variable should be set to enable triggers of the type "action".

diff --git a/.github/workflows/ci.yml b/.github/workflows/ci.yml
@@ -33,7 +33,7 @@ jobs:
           amazon2, amazon2023,
           centos7, centos8, centos9,
           debian8, debian9, debian10, debian11,
-          fedora31, fedora32, fedora34, fedora35, fedora36, fedora37,
+          fedora31, fedora32, fedora34, fedora35, fedora36, fedora37, fedora38,
           ubuntu2004, ubuntu2204
         ]
         arch: [ amd64 ]
@@ -54,6 +54,8 @@ jobs:
             arch: arm64
           - distro: fedora37
             arch: arm64
+          - distro: fedora38
+            arch: arm64
           - distro: ubuntu2204
             arch: arm64
 
@@ -186,15 +188,15 @@ jobs:
             vagrant ssh ${{env.INSTANCE_NAME}} -c "cd tests && sudo ./elio-test.sh -f $fs"
           done
         working-directory: ${{env.BOX_DIR}}
-        # For now tests are taking 10-20 seconds per 1 fs on amd64 and about 2 minutes on arm64. But they can hang.
-        # 10 minutes seems to be reasonable timeout.
-        timeout-minutes: 10
+        # For now tests are taking up to 40 seconds per 1 fs on amd64 and about 2 minutes on arm64 or older 3.x kernels. But they can hang.
+        # 20 minutes seems to be reasonable timeout.
+        timeout-minutes: 20
 
       - name: Show debug information
         if: always()
         run: vagrant ssh ${{env.INSTANCE_NAME}} -c "cat tests/dmesg.log; sudo dmesg -c; lsmod"
         working-directory: ${{env.BOX_DIR}}
-        timeout-minutes: 1
+        timeout-minutes: 2
 
       - name: Run tests on LVM (loop device)
         if: "${{ matrix.distro != 'debian8'}}"
@@ -204,14 +206,14 @@ jobs:
           done
         working-directory: ${{env.BOX_DIR}}
         # For now tests are taking 10-20 seconds per 1 fs on amd64 and about 2 minutes on arm64. But they can hang.
-        # 10 minutes seems to be reasonable timeout.
-        timeout-minutes: 10
+        # 20 minutes seems to be reasonable timeout.
+        timeout-minutes: 20
 
       - name: Show debug information
         if: always()
         run: vagrant ssh ${{env.INSTANCE_NAME}} -c "cat tests/dmesg.log; sudo dmesg -c; lsmod"
         working-directory: ${{env.BOX_DIR}}
-        timeout-minutes: 1
+        timeout-minutes: 2
 
       - name: Run tests on RAID (loop device)
         if: "${{ matrix.distro != 'debian8'}}"
@@ -220,9 +222,9 @@ jobs:
             vagrant ssh ${{env.INSTANCE_NAME}} -c "cd tests && sudo ./elio-test.sh -f $fs --raid"
           done
         working-directory: ${{env.BOX_DIR}}
-        # For now tests are taking 10-20 seconds per 1 fs on amd64 and about 2 minutes on arm64. But they can hang.
-        # 10 minutes seems to be reasonable timeout.
-        timeout-minutes: 10
+        # For now tests are taking up to 40 seconds per 1 fs on amd64 and about 2 minutes on arm64 or older 3.x kernels. But they can hang.
+        # 20 minutes seems to be reasonable timeout.
+        timeout-minutes: 20
 
       - name: Show debug information
         if: always()
@@ -232,10 +234,10 @@ jobs:
 
       - name: Attach qcow2 disks
         run: |
+          SZ="2200M"
           ARCH=$(uname -m)
           TEST_IMAGES=(${TEST_IMAGES})
           TEST_DRIVES=(${TEST_DRIVES})
-          [ ${ARCH} == "x86_64" ] && SZ="1G" || SZ="256M"
           [ ${ARCH} != "x86_64" ] && VIRSH_FLAGS="--config" || true
           for i in ${!TEST_IMAGES[*]}; do
             qemu-img create -f qcow2 ${TEST_IMAGES[i]} $SZ
@@ -261,13 +263,13 @@ jobs:
             vagrant ssh ${{env.INSTANCE_NAME}} -c "cd tests && sudo ./elio-test.sh -d /dev/vdb1 -f $fs"
           done
         working-directory: ${{env.BOX_DIR}}
-        timeout-minutes: 10
+        timeout-minutes: 25
 
       - name: Show debug information
         if: always()
         run: vagrant ssh ${{env.INSTANCE_NAME}} -c "cat tests/dmesg.log; sudo dmesg -c; lsmod"
         working-directory: ${{env.BOX_DIR}}
-        timeout-minutes: 1
+        timeout-minutes: 2
 
       - name: Run tests on LVM (qcow2 disks)
         if: "${{ matrix.distro != 'debian8'}}"
@@ -276,13 +278,13 @@ jobs:
             vagrant ssh ${{env.INSTANCE_NAME}} -c "cd tests && sudo ./elio-test.sh -d /dev/vdb -d /dev/vdc -f $fs --lvm"
           done
         working-directory: ${{env.BOX_DIR}}
-        timeout-minutes: 15
+        timeout-minutes: 25
 
       - name: Show debug information
         if: always()
         run: vagrant ssh ${{env.INSTANCE_NAME}} -c "cat tests/dmesg.log; sudo dmesg -c; lsmod"
         working-directory: ${{env.BOX_DIR}}
-        timeout-minutes: 1
+        timeout-minutes: 2
 
       - name: Run tests on RAID (qcow2 disks)
         if: "${{ matrix.distro != 'debian8'}}"
@@ -294,27 +296,27 @@ jobs:
             vagrant ssh ${{env.INSTANCE_NAME}} -c "cd tests && sudo ./elio-test.sh -d /dev/vdb -d /dev/vdc -f $fs --raid"
           done
         working-directory: ${{env.BOX_DIR}}
-        timeout-minutes: 15
+        timeout-minutes: 25
 
       - name: Show debug information
         if: always()
         run: vagrant ssh ${{env.INSTANCE_NAME}} -c "cat tests/dmesg.log; sudo dmesg -c; lsmod"
         working-directory: ${{env.BOX_DIR}}
-        timeout-minutes: 1
+        timeout-minutes: 2
 
       - name: Run tests multipart (qcow2 disk)
         run: |
           for fs in ${FS[*]}; do
             vagrant ssh ${{env.INSTANCE_NAME}} -c "cd tests && sudo ./elio-test.sh -d /dev/vdb -f $fs -t test_multipart"
           done
         working-directory: ${{env.BOX_DIR}}
-        timeout-minutes: 10
+        timeout-minutes: 15
 
       - name: Show debug information
         if: always()
         run: vagrant ssh ${{env.INSTANCE_NAME}} -c "cat tests/dmesg.log; sudo dmesg -c; lsmod"
         working-directory: ${{env.BOX_DIR}}
-        timeout-minutes: 1
+        timeout-minutes: 2
 
       - name: Detach external drives
         if: always()

diff --git a/src/configure-tests/feature-tests/bio_set_op_attrs.c b/src/configure-tests/feature-tests/bio_set_op_attrs.c
@@ -0,0 +1,16 @@
+// SPDX-License-Identifier: GPL-2.0-only
+
+/*
+ * Copyright (C) 2023 Elastio Software Inc.
+ */
+
+// kernel_version < 6.2
+
+#include "includes.h"
+MODULE_LICENSE("GPL");
+
+static inline void dummy(void){
+	struct bio bio;
+
+	bio_set_op_attrs(&bio, REQ_OP_READ, 0);
+}
diff --git a/src/elastio-snap.c b/src/elastio-snap.c
@@ -298,8 +298,12 @@ typedef enum req_opf req_op_t;
 #endif
 
 static inline void elastio_snap_set_bio_ops(struct bio *bio, req_op_t op, unsigned op_flags){
+#ifdef HAVE_BIO_SET_OP_ATTRS
 	bio->bi_opf = 0;
 	bio_set_op_attrs(bio, op, op_flags);
+#else
+	bio->bi_opf = op | op_flags;
+#endif
 }
 
 static inline int elastio_snap_bio_op_flagged(struct bio *bio, unsigned int flag){
@@ -3341,6 +3345,7 @@ static inline struct inode *page_get_inode(struct page *pg){
 #endif
 	if(PageAnon(pg)) return NULL;
 	if(!pg->mapping) return NULL;
+	if (!virt_addr_valid(pg->mapping)) return NULL;
 	return pg->mapping->host;
 }
 

diff --git a/tests/devicetestcase.py b/tests/devicetestcase.py
@@ -22,7 +22,7 @@ def setUpClass(cls):
         seeds = []
         cls.backing_stores = []
         cls.devices = []
-        cls.size_mb = 256
+        cls.size_mb = 512
         cls.is_raid = False
 
         cls.kmod = kmod.Module("../src/elastio-snap.ko")

diff --git a/tests/devicetestcase_multipart.py b/tests/devicetestcase_multipart.py
@@ -25,7 +25,7 @@ def setUpClass(cls):
 
         # We create the device big enough to perform
         # representative 'test_multipart_modify_origins'
-        cls.size_mb = 1024
+        cls.size_mb = 2200
 
         for i in range(cls.part_count):
             # Unexpectedly randint can generate 2 same numbers in a row.

diff --git a/tests/test_multipart.py b/tests/test_multipart.py
@@ -93,7 +93,7 @@ def test_multipart_modify_origins(self):
 
             # We subtract a couple of megabytes to make sure the cow
             # file won't overflow during the test
-            file_size_mb = math.floor(dev_size_mb * 0.1) - 2
+            file_size_mb = math.floor(dev_size_mb * 0.1) - 20
 
             testfile = "{}/testfile".format(self.mounts[i])
             snapfile = "{}/testfile".format(self.snap_mounts[i])

diff --git a/tests/test_snapshot.py b/tests/test_snapshot.py
@@ -42,7 +42,7 @@ def test_modify_origin(self):
         if self.is_raid == True:
             file_size_mb = math.floor(dev_size_mb * 0.06)
         else:
-            file_size_mb = math.floor(dev_size_mb * 0.1) - 5
+            file_size_mb = math.floor(dev_size_mb * 0.1) - 20
 
         testfile = "{}/testfile".format(self.mount)
         snapfile = "{}/testfile".format(self.snap_mount)

diff --git a/tests/util.py b/tests/util.py
@@ -24,7 +24,7 @@ def unmount(path, retry_on_dev_busy=True):
     cmd = ["umount", path]
     # subprocess.run is introduced in Python 3.5
     if not retry_on_dev_busy or sys.version_info <= (3, 5):
-        subprocess.check_call(cmd, timeout=10)
+        subprocess.check_call(cmd, timeout=20)
     else:
         # The retries on device busy error are necessary on Ubuntu 22.04, kernel 5.15
         # for the tests test_destroy_unverified_incremental and test_destroy_unverified_snapshot.
@@ -48,7 +48,7 @@ def dd(ifile, ofile, count, **kwargs):
     for k, v in kwargs.items():
         cmd.append("{}={}".format(k, v))
 
-    subprocess.check_call(cmd, timeout=60)
+    subprocess.check_call(cmd, timeout=240)
 
 
 def md5sum(path):