| 1 |
On Mon, Oct 14, 2013 at 4:03 PM, David Leverton |
| 2 |
<levertond@××××××××××.com> wrote: |
| 3 |
> Rich Freeman wrote: |
| 4 |
>> |
| 5 |
>> However, FWIW, linux namespaces cannot be used to have only a single |
| 6 |
>> file appear differently to different processes. Mount namespaces can |
| 7 |
>> only operate at the directory level. |
| 8 |
> |
| 9 |
> |
| 10 |
> So to work around that limitation we insist that everyone change how their |
| 11 |
> systems are set up, and still have to reintroduce mtab under a different |
| 12 |
> name ("utab", hidden away under /run) because /proc/self/mounts *doesn't* |
| 13 |
> contain everything that's supposed to be in mtab after all? |
| 14 |
> |
| 15 |
> If someone decides they want to use, say, different DNS servers in different |
| 16 |
> namespaces, should we make the kernel store the server IP addresses, add a |
| 17 |
> /proc file that dumps them out in the expected format, and demand that |
| 18 |
> everyone replace their /etc/resolv.conf with a symlink to |
| 19 |
> /proc/self/resolv.conf? Or maybe, if people want namespaces, they can |
| 20 |
> implement them properly, in which case it becomes literally a self-solving |
| 21 |
> problem. |
| 22 |
> |
| 23 |
> |
| 24 |
|
| 25 |
This is a horrible example. /etc/resolv.conf is a configuration file |
| 26 |
for code that lives entirely in userspace. Of course it makes no sense |
| 27 |
to shove that into the kernel. |
Archives