Bug#854400: systemd: FTBFS if /tmp is an overlayfs

[James Cowgill]

Hi,

On 06/02/17 17:36, Michael Biebl wrote:
> Am 06.02.2017 um 18:07 schrieb James Cowgill:
>> Source: systemd
>> Version: 232-15
>> Severity: important
>>
>> Hi,
>>
>> While testing #852811 I discovered that systemd does not build within an
>> overlayfs schroot (which all my mips chroots are setup as). Specifically
>> the "test-copy" test fails with:
>>
>>> FAIL: test-copy
>>> ===============
>>>
>>> test_copy_file
>>> test_copy_file_fd
>>> test_copy_tree
>>> Assertion 'access(f, F_OK) == 0' failed at ../src/test/test-copy.c:136, function test_copy_tree(). Aborting.
>>> FAIL test-copy (exit status: 134)
>>
>> The assertion is caused by a file not being copied by copy_tree and that
>> is in turn caused by this line:
>> https://sources.debian.net/src/systemd/232-15/src/basic/copy.c/#L342
>>
>> The above line does not copy files if the device of the file is not the
>> same as the device of the parent directory, however on an overlayfs
>> filesystem this is condition is always true for non-directories.
>>
>> From https://www.kernel.org/doc/Documentation/filesystems/overlayfs.txt:
>>> This approach is 'hybrid' because the objects that appear in the
>>> filesystem do not all appear to belong to that filesystem.  In many
>>> cases an object accessed in the union will be indistinguishable
>>> from accessing the corresponding object from the original filesystem.
>>> This is most obvious from the 'st_dev' field returned by stat(2).
>>>
>>> While directories will report an st_dev from the overlay-filesystem,
>>> all non-directory objects will report an st_dev from the lower or
>>> upper filesystem that is providing the object.  Similarly st_ino will
>>> only be unique when combined with st_dev, and both of these can change
>>> over the lifetime of a non-directory object.  Many applications and
>>> tools ignore these values and will not be affected.
>>
>> Note that this behavior is different from say 'cp -Rx'. cp only checks
>> the value of st_dev if the file is a directory. This allows it to work
>> correctly with overlayfs.
>> http://sources.debian.net/src/coreutils/8.26-2/src/copy.c/?hl=2547#L2547
> 
> I think this is a bug in overlayfs.

Well it's the documented behavior of overlayfs which others have
probably relied on, so I think it's very unlikely overlayfs will be
changed now.

James

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 833 bytes
Desc: OpenPGP digital signature
URL: <http://alioth-lists.debian.net/pipermail/pkg-systemd-maintainers/attachments/20170206/9925f528/attachment-0002.sig>