Quoting Oren Laadan (or...@cs.columbia.edu):
> We only allow c/r when all processes shared a single mounts ns.
>
> We do intend to implement c/r of mounts and mounts namespaces in the
> kernel. It shouldn't be ugly or complicate locking to do so. Just
> haven't gotten around to it. A more comple
Quoting Oren Laadan (or...@cs.columbia.edu):
> When this new switch is specified, the coordinator will unshare the
> mounts namespace and restart will occur inside a private namespace.
> Without it the restarting tasks will share the current mounts
> namespace.
>
> Signed-off-by: Oren Laadan
Ack
Quoting Oren Laadan (or...@cs.columbia.edu):
> Checkpoint and restore task->fs. Tasks sharing task->fs will
> share them again after restart.
>
> Original patch by Serge Hallyn
>
> Changelog:
> Jan 25: [orenl] Addressed comments by .. myself:
> - add leak detection
> - change order of
When this new switch is specified, the coordinator will unshare the
mounts namespace and restart will occur inside a private namespace.
Without it the restarting tasks will share the current mounts
namespace.
Signed-off-by: Oren Laadan
---
restart.c | 16 ++--
1 files changed, 14 i
We only allow c/r when all processes shared a single mounts ns.
We do intend to implement c/r of mounts and mounts namespaces in the
kernel. It shouldn't be ugly or complicate locking to do so. Just
haven't gotten around to it. A more complete solution is more than we
want to take on now for v19
From: Serge E. Hallyn
I'm hoping to get restart of 32-bit tasks on x86-64 working
for v19, but restart of 32-bit tasks using a 64-bit restart
program I'm not going to even try right now. Rather than
get a completely uncommented '-22' return value from restore_cpu()
or do_munmap(), catch it early
From: Serge E. Hallyn
1. always save/restore fs/gs
not just if CONFIG_COMPAT and the task is 64-bit. Pretty
sure that block wasn't doing what it wanted to do anyway :)
2. use rdmsrl for fs/gs for self-checkpoint
other code seems to do it. It doesn't seem to make
From: Serge E. Hallyn
Eclone on x86-32 should use esi for 4th arg (upids), not
edi, which is 5th arg. This will require a corresponding fix
to user-cr, but (a) it's correct and (b) it's required so that
IA32_ARG_FIXUP can put upids in the right register for
process_64.c:sys_eclone(). IA32_ARG_F
The first patch is a resend from friday, and should be applied to
v19. The second patch I think should be applied to v19 as well,
as it will prevent accidental attempts to restart a 64-bit image
with a 32-bit restart, and vice versa.
The third patch I think does things which need to be done, but
Checkpoint and restore task->fs. Tasks sharing task->fs will
share them again after restart.
Original patch by Serge Hallyn
Changelog:
Jan 25: [orenl] Addressed comments by .. myself:
- add leak detection
- change order of save/restore of chroot and cwd
- save/restore fs only afte
10 matches
Mail list logo