[CRIU] [PATCH v4 8/9] mount: do remaps for child-overmount of another overmount

Pavel Tikhomirov ptikhomirov at virtuozzo.com
Thu Sep 27 10:02:05 MSK 2018



On 09/27/2018 12:22 AM, Andrey Vagin wrote:
> On Mon, Dec 11, 2017 at 07:07:25PM +0300, Pavel Tikhomirov wrote:
>> In case we have mounts:
>>
>> 1 /mnt/
>> 2 /mnt/a with parent 1
>> 3 /mnt/a/b with parent 1
>> 4 /mnt/a with parent 2
>>
>> We determine 2 as needing remap with does_mnt_overmount() and remap it.
> 
> Why do we determine that 2 has to be remmapped? It overmounts nothing.

2 is a sibling-overmount for 3, that's why

> 
> The test passes without this patch. Could you explain why do we need
> this patch? I think we need to improve the test to check the cases what
> is fixed by this patch.

For me if I revert the patch, test fails with lost mount. (If we don't 
remap 4, the mount 2 will be lost.

Run criu restore
1: Old mounts lost: [(u'/', u'/zdtm/static/tempfs_overmounted.test/a', 
None, None)]
1: New mounts appeared: []
########## Test zdtm/static/tempfs_overmounted FAIL at mounts compare 
##########

> 
> 
>> Next we mount 4 on top of 2. Next in fixup_remap_mounts() we want to
>> move 2 back to it's parent 1, but instead move 4 there. So in these case
>> children-overmounts need to be remapped too.
>>
>> Signed-off-by: Pavel Tikhomirov <ptikhomirov at virtuozzo.com>
>> ---
>>   criu/mount.c | 12 ++++++++++--
>>   1 file changed, 10 insertions(+), 2 deletions(-)
>>
>> diff --git a/criu/mount.c b/criu/mount.c
>> index bacb24222..cdf1c0432 100644
>> --- a/criu/mount.c
>> +++ b/criu/mount.c
>> @@ -630,7 +630,7 @@ static struct mount_info *find_fsroot_mount_for(struct mount_info *bm)
>>   	return NULL;
>>   }
>>   
>> -static bool does_mnt_overmount(struct mount_info *m)
>> +static bool mnt_needs_remap(struct mount_info *m)
>>   {
>>   	struct mount_info *t;
>>   
>> @@ -644,6 +644,14 @@ static bool does_mnt_overmount(struct mount_info *m)
>>   			return true;
>>   	}
>>   
>> +	/*
>> +	 * If we are children-overmount and parent is remapped, we should be
>> +	 * remapped too, else fixup_remap_mounts() won't be able to move parent
>> +	 * to it's real place, it will move child instead.
>> +	 */
>> +	if (!strcmp(m->parent->mountpoint, m->mountpoint))
>> +		return mnt_needs_remap(m->parent);
>> +
>>   	return false;
>>   }
>>   
>> @@ -2488,7 +2496,7 @@ static int try_remap_mount(struct mount_info *m)
>>   {
>>   	struct mnt_remap_entry *r;
>>   
>> -	if (!does_mnt_overmount(m))
>> +	if (!mnt_needs_remap(m))
>>   		return 0;
>>   
>>   	BUG_ON(!m->parent);
>> -- 
>> 2.14.3
>>

-- 
Best regards, Tikhomirov Pavel
Software Developer, Virtuozzo.


More information about the CRIU mailing list