Skip to contentSkip to conversation

test/py: Use loop mounts if guestmount fails in filesystem tests

PATCH In review Alper Nebi Yasak 3 participants 8 messages Jun 4, 2021 – Jul 6, 2021

lore
Share & tools
lore
#0 lore raw

PATCH 1/2 test/py: Use loop mounts if guestmount fails in filesystem tests

If guestmount isn't available on the system, filesystem test setup falls back to using loop mounts to prepare its disk images. If guestmount is available but fails to work, the tests are immediately skipped. Instead of giving up on a guestmount failure, try using loop mounts as an attempt to keep tests running.

Also stop checking if guestmount is in PATH, as trying to run a missing guestmount can now follow the same failure codepath and fall back to loop mounts anyway.

---

 test/py/tests/test_fs/conftest.py | 31 +++++++++++++++----------------
 1 file changed, 15 insertions(+), 16 deletions(-)
test/py/tests/test_fs/conftest.py +15 -16
index 410a675b9714..e3c461635f8e 100644
@@ -205,24 +205,23 @@ def mount_fs(fs_type, device, mount_point):
     """
     global fuse_mounted
 
-    fuse_mounted = False
     try:
-        if tool_is_in_path('guestmount'):
-            fuse_mounted = True
-            check_call('guestmount -a %s -m /dev/sda %s'
-                % (device, mount_point), shell=True)
-        else:
-            mount_opt = 'loop,rw'
-            if re.match('fat', fs_type):
-                mount_opt += ',umask=0000'
-
-            check_call('sudo mount -o %s %s %s'
-                % (mount_opt, device, mount_point), shell=True)
-
-            # may not be effective for some file systems
-            check_call('sudo chmod a+rw %s' % mount_point, shell=True)
+        check_call('guestmount -a %s -m /dev/sda %s'
+            % (device, mount_point), shell=True)
+        fuse_mounted = True
+        return
     except CalledProcessError:
-        raise
+        fuse_mounted = False
+
+    mount_opt = 'loop,rw'
+    if re.match('fat', fs_type):
+        mount_opt += ',umask=0000'
+
+    check_call('sudo mount -o %s %s %s'
+        % (mount_opt, device, mount_point), shell=True)
+
+    # may not be effective for some file systems
+    check_call('sudo chmod a+rw %s' % mount_point, shell=True)
 
 def umount_fs(mount_point):
     """Unmount a volume.
Signature · 1 line
2.32.0.rc2
#1 lore raw

PATCH 2/2 test/py: Wait for guestmount worker to exit after running guestunmount

Some filesystem tests are failing when their image is prepared with guestmount, but succeeding if loop mounts are used instead. The reason seems to be a race condition the guestmount(1) manual page explains:

When guestunmount(1)/fusermount(1) exits, guestmount may still be running and cleaning up the mountpoint. The disk image will not be fully finalized.

This means that scripts like the following have a nasty race condition:

guestmount -a disk.img -i /mnt
# copy things into /mnt
guestunmount /mnt
# immediately try to use 'disk.img' ** UNSAFE **

The solution is to use the --pid-file option to write the guestmount PID to a file, then after guestunmount spin waiting for this PID to exit.

The Python standard library has an os.waitpid() function for waiting a child to terminate, but it cannot wait on non-child processes. Implement a utility function that can do this by polling the process repeatedly for a given duration, optionally killing the process if it won't terminate on its own. Apply the suggested solution with this utility function, which makes the failing tests succeed again.

---

 test/py/tests/test_fs/conftest.py | 13 ++++++++++-
 test/py/u_boot_utils.py           | 36 +++++++++++++++++++++++++++++++
 2 files changed, 48 insertions(+), 1 deletion(-)
test/py/tests/test_fs/conftest.py +12 -1
index e3c461635f8e..6b1ff05a8143 100644
@@ -8,6 +8,7 @@
 import re
 from subprocess import call, check_call, check_output, CalledProcessError
 from fstest_defs import *
+import u_boot_utils as util
 
 supported_fs_basic = ['fat16', 'fat32', 'ext4']
 supported_fs_ext = ['fat16', 'fat32']
@@ -206,7 +207,7 @@ def mount_fs(fs_type, device, mount_point):
     global fuse_mounted
 
     try:
-        check_call('guestmount -a %s -m /dev/sda %s'
+        check_call('guestmount --pid-file guestmount.pid -a %s -m /dev/sda %s'
             % (device, mount_point), shell=True)
         fuse_mounted = True
         return
@@ -235,6 +236,16 @@ def umount_fs(mount_point):
     if fuse_mounted:
         call('sync')
         call('guestunmount %s' % mount_point, shell=True)
+
+        try:
+            with open("guestmount.pid", "r") as pidfile:
+                pid = int(pidfile.read())
+            util.waitpid(pid, kill=True)
+            os.remove("guestmount.pid")
+
+        except FileNotFoundError:
+            pass
+
     else:
         call('sudo umount %s' % mount_point, shell=True)
 
test/py/u_boot_utils.py +36 -0
index 939d82eec12a..e816c7fbb6a3 100644
@@ -8,6 +8,7 @@
 import os
 import os.path
 import pytest
+import signal
 import sys
 import time
 import re
@@ -339,3 +340,38 @@ def crc32(u_boot_console, address, count):
     assert m, 'CRC32 operation failed.'
 
     return m.group(1)
+
+def waitpid(pid, timeout=60, kill=False):
+    """Wait a process to terminate by its PID
+
+    This is an alternative to a os.waitpid(pid, 0) call that works on
+    processes that aren't children of the python process.
+
+    Args:
+        pid: PID of a running process.
+        timeout: Time in seconds to wait.
+        kill: Whether to forcibly kill the process after timeout.
+
+    Returns:
+        True, if the process ended on its own.
+        False, if the process was killed by this function.
+
+    Raises:
+        TimeoutError, if the process is still running after timeout.
+    """
+    try:
+        for _ in range(timeout):
+            os.kill(pid, 0)
+            time.sleep(1)
+
+        if kill:
+            os.kill(pid, signal.SIGKILL)
+            return False
+
+    except ProcessLookupError:
+        return True
+
+    raise TimeoutError(
+        "Process with PID {} did not terminate after {} seconds."
+        .format(pid, timeout)
+    )
Signature · 1 line
2.32.0.rc2
Reviewed-by on v1 2/2
#2 lore raw

Re: [PATCH 2/2] test/py: Wait for guestmount worker to exit after running guestunmount

Quoted 27 lines · Alper Nebi Yasak

Some filesystem tests are failing when their image is prepared with guestmount, but succeeding if loop mounts are used instead. The reason seems to be a race condition the guestmount(1) manual page explains:

When guestunmount(1)/fusermount(1) exits, guestmount may still be running and cleaning up the mountpoint. The disk image will not be fully finalized.

This means that scripts like the following have a nasty race condition:

guestmount -a disk.img -i /mnt
# copy things into /mnt
guestunmount /mnt
# immediately try to use 'disk.img' ** UNSAFE **

The solution is to use the --pid-file option to write the guestmount PID to a file, then after guestunmount spin waiting for this PID to exit.

The Python standard library has an os.waitpid() function for waiting a child to terminate, but it cannot wait on non-child processes. Implement a utility function that can do this by polling the process repeatedly for a given duration, optionally killing the process if it won't terminate on its own. Apply the suggested solution with this utility function, which makes the failing tests succeed again.

Quoted 7 lines

---

 test/py/tests/test_fs/conftest.py | 13 ++++++++++-
 test/py/u_boot_utils.py           | 36 +++++++++++++++++++++++++++++++
 2 files changed, 48 insertions(+), 1 deletion(-)
Quoted 22 lines
test/py/tests/test_fs/conftest.py +2 -1
index e3c461635f8e..6b1ff05a8143 100644
@@ -8,6 +8,7 @@
 import re
 from subprocess import call, check_call, check_output, CalledProcessError
 from fstest_defs import *
+import u_boot_utils as util
 
 supported_fs_basic = ['fat16', 'fat32', 'ext4']
 supported_fs_ext = ['fat16', 'fat32']
@@ -206,7 +207,7 @@ def mount_fs(fs_type, device, mount_point):
     global fuse_mounted
 
     try:
-        check_call('guestmount -a %s -m /dev/sda %s'
+        check_call('guestmount --pid-file guestmount.pid -a %s -m /dev/sda %s'
             % (device, mount_point), shell=True)
         fuse_mounted = True
         return

test/py/tests/test_fs/conftest.py def umount_fs(mount_point): · line 237

Quoted 2 lines
test/py/tests/test_fs/conftest.py +0 -0
@@ -235,6 +236,16 @@ def umount_fs(mount_point):
     if fuse_mounted:
         call('sync')

should you remove guestmount.pid first in case it is there from a previous crash?

Quoted 68 lines
test/py/tests/test_fs/conftest.py +10 -0
         call('guestunmount %s' % mount_point, shell=True)
+
+        try:
+            with open("guestmount.pid", "r") as pidfile:
+                pid = int(pidfile.read())
+            util.waitpid(pid, kill=True)
+            os.remove("guestmount.pid")
+
+        except FileNotFoundError:
+            pass
+
     else:
         call('sudo umount %s' % mount_point, shell=True)
 
test/py/u_boot_utils.py +36 -1
index 939d82eec12a..e816c7fbb6a3 100644
@@ -8,6 +8,7 @@
 import os
 import os.path
 import pytest
+import signal
 import sys
 import time
 import re
@@ -339,3 +340,38 @@ def crc32(u_boot_console, address, count):
     assert m, 'CRC32 operation failed.'
 
     return m.group(1)
+
+def waitpid(pid, timeout=60, kill=False):
+    """Wait a process to terminate by its PID
+
+    This is an alternative to a os.waitpid(pid, 0) call that works on
+    processes that aren't children of the python process.
+
+    Args:
+        pid: PID of a running process.
+        timeout: Time in seconds to wait.
+        kill: Whether to forcibly kill the process after timeout.
+
+    Returns:
+        True, if the process ended on its own.
+        False, if the process was killed by this function.
+
+    Raises:
+        TimeoutError, if the process is still running after timeout.
+    """
+    try:
+        for _ in range(timeout):
+            os.kill(pid, 0)
+            time.sleep(1)
+
+        if kill:
+            os.kill(pid, signal.SIGKILL)
+            return False
+
+    except ProcessLookupError:
+        return True
+
+    raise TimeoutError(
+        "Process with PID {} did not terminate after {} seconds."
+        .format(pid, timeout)
+    )
--

2.32.0.rc2

#3 lore raw

Re: [PATCH 2/2] test/py: Wait for guestmount worker to exit after running guestunmount

Quoted 53 lines · Simon Glass
Quoted 27 lines · Alper Nebi Yasak

Some filesystem tests are failing when their image is prepared with guestmount, but succeeding if loop mounts are used instead. The reason seems to be a race condition the guestmount(1) manual page explains:

When guestunmount(1)/fusermount(1) exits, guestmount may still be running and cleaning up the mountpoint. The disk image will not be fully finalized.

This means that scripts like the following have a nasty race condition:

guestmount -a disk.img -i /mnt
# copy things into /mnt
guestunmount /mnt
# immediately try to use 'disk.img' ** UNSAFE **

The solution is to use the --pid-file option to write the guestmount PID to a file, then after guestunmount spin waiting for this PID to exit.

The Python standard library has an os.waitpid() function for waiting a child to terminate, but it cannot wait on non-child processes. Implement a utility function that can do this by polling the process repeatedly for a given duration, optionally killing the process if it won't terminate on its own. Apply the suggested solution with this utility function, which makes the failing tests succeed again.

Quoted 7 lines

---

 test/py/tests/test_fs/conftest.py | 13 ++++++++++-
 test/py/u_boot_utils.py           | 36 +++++++++++++++++++++++++++++++
 2 files changed, 48 insertions(+), 1 deletion(-)
Quoted 15 lines
test/py/tests/test_fs/conftest.py +1 -0
index e3c461635f8e..6b1ff05a8143 100644
@@ -8,6 +8,7 @@
 import re
 from subprocess import call, check_call, check_output, CalledProcessError
 from fstest_defs import *
+import u_boot_utils as util
 
 supported_fs_basic = ['fat16', 'fat32', 'ext4']
 supported_fs_ext = ['fat16', 'fat32']
@@ -206,7 +207,7 @@ def mount_fs(fs_type, device, mount_point):
     global fuse_mounted
 
Quoted 12 lines

test/py/tests/test_fs/conftest.py def umount_fs(mount_point): · line 237

Quoted 9 lines
test/py/tests/test_fs/conftest.py +1 -1
     try:
-        check_call('guestmount -a %s -m /dev/sda %s'
+        check_call('guestmount --pid-file guestmount.pid -a %s -m /dev/sda %s'
             % (device, mount_point), shell=True)
         fuse_mounted = True
         return
@@ -235,6 +236,16 @@ def umount_fs(mount_point):
     if fuse_mounted:
         call('sync')

should you remove guestmount.pid first in case it is there from a previous crash?

Guestmount overwrites the pid file without complaining, and the worker deletes the file on its own when it terminates after the guestunmount call.

I think it's OK to ignore an existing pid file if the mount point is clean. If not clean, that's the actual problem and both guestmount and mount will fail, causing the tests to be skipped. We could clean mount points automatically before attempting to mount.

Quoted 68 lines
Quoted 68 lines
test/py/tests/test_fs/conftest.py +10 -0
         call('guestunmount %s' % mount_point, shell=True)
+
+        try:
+            with open("guestmount.pid", "r") as pidfile:
+                pid = int(pidfile.read())
+            util.waitpid(pid, kill=True)
+            os.remove("guestmount.pid")
+
+        except FileNotFoundError:
+            pass
+
     else:
         call('sudo umount %s' % mount_point, shell=True)
 
test/py/u_boot_utils.py +36 -1
index 939d82eec12a..e816c7fbb6a3 100644
@@ -8,6 +8,7 @@
 import os
 import os.path
 import pytest
+import signal
 import sys
 import time
 import re
@@ -339,3 +340,38 @@ def crc32(u_boot_console, address, count):
     assert m, 'CRC32 operation failed.'
 
     return m.group(1)
+
+def waitpid(pid, timeout=60, kill=False):
+    """Wait a process to terminate by its PID
+
+    This is an alternative to a os.waitpid(pid, 0) call that works on
+    processes that aren't children of the python process.
+
+    Args:
+        pid: PID of a running process.
+        timeout: Time in seconds to wait.
+        kill: Whether to forcibly kill the process after timeout.
+
+    Returns:
+        True, if the process ended on its own.
+        False, if the process was killed by this function.
+
+    Raises:
+        TimeoutError, if the process is still running after timeout.
+    """
+    try:
+        for _ in range(timeout):
+            os.kill(pid, 0)
+            time.sleep(1)
+
+        if kill:
+            os.kill(pid, signal.SIGKILL)
+            return False
+
+    except ProcessLookupError:
+        return True
+
+    raise TimeoutError(
+        "Process with PID {} did not terminate after {} seconds."
+        .format(pid, timeout)
+    )
--

2.32.0.rc2

#4 lore raw

Re: [PATCH 2/2] test/py: Wait for guestmount worker to exit after running guestunmount

In reply to Alper Nebi Yasak
Quoted 26 lines · Alper Nebi Yasak

Some filesystem tests are failing when their image is prepared with guestmount, but succeeding if loop mounts are used instead. The reason seems to be a race condition the guestmount(1) manual page explains:

When guestunmount(1)/fusermount(1) exits, guestmount may still be running and cleaning up the mountpoint. The disk image will not be fully finalized.

This means that scripts like the following have a nasty race condition:

guestmount -a disk.img -i /mnt
# copy things into /mnt
guestunmount /mnt
# immediately try to use 'disk.img' ** UNSAFE **

The solution is to use the --pid-file option to write the guestmount PID to a file, then after guestunmount spin waiting for this PID to exit.

The Python standard library has an os.waitpid() function for waiting a child to terminate, but it cannot wait on non-child processes. Implement a utility function that can do this by polling the process repeatedly for a given duration, optionally killing the process if it won't terminate on its own. Apply the suggested solution with this utility function, which makes the failing tests succeed again.

Quoted 2 lines

Applied to u-boot/master, thanks!

Signature · 1 line
Tom
#5 lore raw
In reply to Alper Nebi Yasak
Quoted 11 lines · Alper Nebi Yasak

If guestmount isn't available on the system, filesystem test setup falls back to using loop mounts to prepare its disk images. If guestmount is available but fails to work, the tests are immediately skipped. Instead of giving up on a guestmount failure, try using loop mounts as an attempt to keep tests running.

Also stop checking if guestmount is in PATH, as trying to run a missing guestmount can now follow the same failure codepath and fall back to loop mounts anyway.

Quoted 5 lines

---

 test/py/tests/test_fs/conftest.py | 31 +++++++++++++++----------------
 1 file changed, 15 insertions(+), 16 deletions(-)

https://source.denx.de/u-boot/custodians/u-boot-dm/-/jobs/282788

thoughts below

Quoted 25 lines
test/py/tests/test_fs/conftest.py +0 -15
index 410a675b9714..e3c461635f8e 100644
@@ -205,24 +205,23 @@ def mount_fs(fs_type, device, mount_point):
     """
     global fuse_mounted
 
-    fuse_mounted = False
     try:
-        if tool_is_in_path('guestmount'):
-            fuse_mounted = True
-            check_call('guestmount -a %s -m /dev/sda %s'
-                % (device, mount_point), shell=True)
-        else:
-            mount_opt = 'loop,rw'
-            if re.match('fat', fs_type):
-                mount_opt += ',umask=0000'
-
-            check_call('sudo mount -o %s %s %s'
-                % (mount_opt, device, mount_point), shell=True)
-
-            # may not be effective for some file systems
-            check_call('sudo chmod a+rw %s' % mount_point, shell=True)

test/py/tests/test_fs/conftest.py def mount_fs(fs_type, device, mount_point): · line 217

Quoted 10 lines
test/py/tests/test_fs/conftest.py +8 -1
+        check_call('guestmount -a %s -m /dev/sda %s'
+            % (device, mount_point), shell=True)
+        fuse_mounted = True
+        return
     except CalledProcessError:
-        raise
+        fuse_mounted = False
+
+    mount_opt = 'loop,rw'
+    if re.match('fat', fs_type):

This seems ambiguous. I think you should use:

if 'fat' in fs_type:

or

if fs_type.startswith('fat'):

or

if fs_type == 'fat':

so it is clear what you mean.

Quoted 11 lines
test/py/tests/test_fs/conftest.py +7 -1
+        mount_opt += ',umask=0000'
+
+    check_call('sudo mount -o %s %s %s'
+        % (mount_opt, device, mount_point), shell=True)
+
+    # may not be effective for some file systems
+    check_call('sudo chmod a+rw %s' % mount_point, shell=True)
 
 def umount_fs(mount_point):
     """Unmount a volume.
--
Quoted 2 lines

2.32.0.rc2

Regards,
Simon

#6 lore raw
Quoted 15 lines · Simon Glass
Quoted 14 lines · Alper Nebi Yasak

If guestmount isn't available on the system, filesystem test setup falls back to using loop mounts to prepare its disk images. If guestmount is available but fails to work, the tests are immediately skipped. Instead of giving up on a guestmount failure, try using loop mounts as an attempt to keep tests running.

Also stop checking if guestmount is in PATH, as trying to run a missing guestmount can now follow the same failure codepath and fall back to loop mounts anyway.

---

Quoted 6 lines
Quoted 2 lines
 test/py/tests/test_fs/conftest.py | 31 +++++++++++++++----------------
 1 file changed, 15 insertions(+), 16 deletions(-)

https://source.denx.de/u-boot/custodians/u-boot-dm/-/jobs/282788

thoughts below

That CI run doesn't have this patch/series applied.

Quoted 42 lines
Quoted 25 lines
test/py/tests/test_fs/conftest.py +0 -15
index 410a675b9714..e3c461635f8e 100644
@@ -205,24 +205,23 @@ def mount_fs(fs_type, device, mount_point):
     """
     global fuse_mounted
 
-    fuse_mounted = False
     try:
-        if tool_is_in_path('guestmount'):
-            fuse_mounted = True
-            check_call('guestmount -a %s -m /dev/sda %s'
-                % (device, mount_point), shell=True)
-        else:
-            mount_opt = 'loop,rw'
-            if re.match('fat', fs_type):
-                mount_opt += ',umask=0000'
-
-            check_call('sudo mount -o %s %s %s'
-                % (mount_opt, device, mount_point), shell=True)
-
-            # may not be effective for some file systems
-            check_call('sudo chmod a+rw %s' % mount_point, shell=True)

test/py/tests/test_fs/conftest.py def mount_fs(fs_type, device, mount_point): · line 217

Quoted 10 lines
test/py/tests/test_fs/conftest.py +8 -1
+        check_call('guestmount -a %s -m /dev/sda %s'
+            % (device, mount_point), shell=True)
+        fuse_mounted = True
+        return
     except CalledProcessError:
-        raise
+        fuse_mounted = False
+
+    mount_opt = 'loop,rw'
+    if re.match('fat', fs_type):

This seems ambiguous. I think you should use:

if 'fat' in fs_type:

or

Quoted 7 lines

if fs_type.startswith('fat'):

or

if fs_type == 'fat':

so it is clear what you mean.

This just unindents existing code. I can send v2 with startswith('fat') though.

(I think I should rewrite the tests with guestfish and remove these mount functions instead; but I'm busy with other things for a while.)

Quoted 17 lines
Quoted 11 lines
test/py/tests/test_fs/conftest.py +7 -1
+        mount_opt += ',umask=0000'
+
+    check_call('sudo mount -o %s %s %s'
+        % (mount_opt, device, mount_point), shell=True)
+
+    # may not be effective for some file systems
+    check_call('sudo chmod a+rw %s' % mount_point, shell=True)
 
 def umount_fs(mount_point):
     """Unmount a volume.
--
Quoted 2 lines

2.32.0.rc2

Regards,
Simon

#7 lore raw
In reply to Alper Nebi Yasak
Quoted 11 lines · Alper Nebi Yasak

If guestmount isn't available on the system, filesystem test setup falls back to using loop mounts to prepare its disk images. If guestmount is available but fails to work, the tests are immediately skipped. Instead of giving up on a guestmount failure, try using loop mounts as an attempt to keep tests running.

Also stop checking if guestmount is in PATH, as trying to run a missing guestmount can now follow the same failure codepath and fall back to loop mounts anyway.

Applied to u-boot/master, thanks!

Signature · 1 line
Tom