From: Cedric Le Goater <[email protected]>
We have at least 2 patchsets requiring each a new clone flag and
there it is, we've reached the limit, none are left.
This patch uses the CLONE_DETACHED flag (unused) as a marker to
extend the clone flags through the parent_tidptr argument.
Initially, we thought on using the last bit but it has recently
been taken by CLONE_IO.
Obviously, this hack doesn't work for unshare() for which I don't
see any other solution than to add a new syscall :
long sys_unshare64(unsigned long clone_flags_high,
unsigned long clone_flags_low);
Is this the right path to extend the clone flags ? should we add a
clone64() rather than hack the extending clone() ?
Thanks for any comments !
C.
Signed-off-by: Cedric Le Goater <[email protected]>
---
include/linux/sched.h | 1 +
kernel/fork.c | 14 +++++++++++++-
2 files changed, 14 insertions(+), 1 deletion(-)
Index: 2.6.24-mm1/include/linux/sched.h
===================================================================
--- 2.6.24-mm1.orig/include/linux/sched.h
+++ 2.6.24-mm1/include/linux/sched.h
@@ -28,6 +28,7 @@
#define CLONE_NEWPID 0x20000000 /* New pid namespace */
#define CLONE_NEWNET 0x40000000 /* New network namespace */
#define CLONE_IO 0x80000000 /* Clone io context */
+#define CLONE_EXTFLAGS CLONE_DETACHED /* use parent_tidptr as an extended set of flags */
/*
* Scheduling policies
Index: 2.6.24-mm1/kernel/fork.c
===================================================================
--- 2.6.24-mm1.orig/kernel/fork.c
+++ 2.6.24-mm1/kernel/fork.c
@@ -1012,6 +1012,14 @@ static struct task_struct *copy_process(
struct task_struct *p;
int cgroup_callbacks_done = 0;
+ /*
+ * It is not permitted to specify both CLONE_EXTFLAGS and
+ * CLONE_PARENT_SETTID
+ */
+ if ((clone_flags & (CLONE_EXTFLAGS|CLONE_PARENT_SETTID)) ==
+ (CLONE_EXTFLAGS|CLONE_PARENT_SETTID))
+ return ERR_PTR(-EINVAL);
+
if ((clone_flags & (CLONE_NEWNS|CLONE_FS)) == (CLONE_NEWNS|CLONE_FS))
return ERR_PTR(-EINVAL);
@@ -1455,6 +1463,7 @@ long do_fork(unsigned long clone_flags,
struct task_struct *p;
int trace = 0;
long nr;
+ u64 clone_flags64 = clone_flags;
/*
* We hope to recycle these flags after 2.6.26
@@ -1479,7 +1488,10 @@ long do_fork(unsigned long clone_flags,
clone_flags |= CLONE_PTRACE;
}
- p = copy_process(clone_flags, stack_start, regs, stack_size,
+ if (clone_flags & CLONE_EXTFLAGS)
+ clone_flags64 = ((u64) (uintptr_t) parent_tidptr << 32) | clone_flags;
+
+ p = copy_process(clone_flags64, stack_start, regs, stack_size,
child_tidptr, NULL);
/*
* Do this prior waking up the new thread - the thread pointer
Quoting Cedric Le Goater ([email protected]):
> From: Cedric Le Goater <[email protected]>
>
> We have at least 2 patchsets requiring each a new clone flag and there it
> is, we've reached the limit, none are left.
> This patch uses the CLONE_DETACHED flag (unused) as a marker to extend the
Are we pretty sure that there is no legacy software out there which has
continued to specify CLONE_DETACHED since the kernel ignores it?
> clone flags through the parent_tidptr argument.
>
> Initially, we thought on using the last bit but it has recently been taken
> by CLONE_IO.
>
> Obviously, this hack doesn't work for unshare() for which I don't see any
> other solution than to add a new syscall :
> long sys_unshare64(unsigned long clone_flags_high, unsigned long
> clone_flags_low);
>
>
>
> Is this the right path to extend the clone flags ? should we add a
> clone64() rather than hack the extending clone() ?
> Thanks for any comments !
>
> C.
>
> Signed-off-by: Cedric Le Goater <[email protected]>
> ---
> include/linux/sched.h | 1 +
> kernel/fork.c | 14 +++++++++++++-
> 2 files changed, 14 insertions(+), 1 deletion(-)
>
> Index: 2.6.24-mm1/include/linux/sched.h
> ===================================================================
> --- 2.6.24-mm1.orig/include/linux/sched.h
> +++ 2.6.24-mm1/include/linux/sched.h
> @@ -28,6 +28,7 @@
> #define CLONE_NEWPID 0x20000000 /* New pid namespace */
> #define CLONE_NEWNET 0x40000000 /* New network namespace */
> #define CLONE_IO 0x80000000 /* Clone io context */
> +#define CLONE_EXTFLAGS CLONE_DETACHED /* use parent_tidptr as an extended
> set of flags */
>
> /*
> * Scheduling policies
> Index: 2.6.24-mm1/kernel/fork.c
> ===================================================================
> --- 2.6.24-mm1.orig/kernel/fork.c
> +++ 2.6.24-mm1/kernel/fork.c
> @@ -1012,6 +1012,14 @@ static struct task_struct *copy_process(
> struct task_struct *p;
> int cgroup_callbacks_done = 0;
>
> + /*
> + * It is not permitted to specify both CLONE_EXTFLAGS and
> + * CLONE_PARENT_SETTID
> + */
> + if ((clone_flags & (CLONE_EXTFLAGS|CLONE_PARENT_SETTID)) ==
> + (CLONE_EXTFLAGS|CLONE_PARENT_SETTID))
> + return ERR_PTR(-EINVAL);
> +
> if ((clone_flags & (CLONE_NEWNS|CLONE_FS)) == (CLONE_NEWNS|CLONE_FS))
> return ERR_PTR(-EINVAL);
>
> @@ -1455,6 +1463,7 @@ long do_fork(unsigned long clone_flags,
> struct task_struct *p;
> int trace = 0;
> long nr;
> + u64 clone_flags64 = clone_flags;
>
> /*
> * We hope to recycle these flags after 2.6.26
> @@ -1479,7 +1488,10 @@ long do_fork(unsigned long clone_flags,
> clone_flags |= CLONE_PTRACE;
> }
>
> - p = copy_process(clone_flags, stack_start, regs, stack_size,
> + if (clone_flags & CLONE_EXTFLAGS)
> + clone_flags64 = ((u64) (uintptr_t) parent_tidptr << 32) | clone_flags;
> +
> + p = copy_process(clone_flags64, stack_start, regs, stack_size,
> child_tidptr, NULL);
> /*
> * Do this prior waking up the new thread - the thread pointer
> _______________________________________________
> Containers mailing list
> [email protected]
> https://lists.linux-foundation.org/mailman/listinfo/containers
On Mon, 4 Feb 2008 14:24:16 -0600
"Serge E. Hallyn" <[email protected]> wrote:
> Quoting Cedric Le Goater ([email protected]):
> > From: Cedric Le Goater <[email protected]>
> >
> > We have at least 2 patchsets requiring each a new clone flag and there it
> > is, we've reached the limit, none are left.
> > This patch uses the CLONE_DETACHED flag (unused) as a marker to extend the
>
> Are we pretty sure that there is no legacy software out there which has
> continued to specify CLONE_DETACHED since the kernel ignores it?
Please see -mm's
clone-prepare-to-recycle-clone_detached-and-clone_stopped.patch
That patch has been cut back to only recycle CLONE_STOPPED because there
indeed was software out there which is setting CLONE_DETACHED.
See http://linux.derkeiler.com/Mailing-Lists/Kernel/2007-11/msg04293.html
Andrew Morton wrote:
> On Mon, 4 Feb 2008 14:24:16 -0600
> "Serge E. Hallyn" <[email protected]> wrote:
>
>> Quoting Cedric Le Goater ([email protected]):
>>> From: Cedric Le Goater <[email protected]>
>>>
>>> We have at least 2 patchsets requiring each a new clone flag and there it
>>> is, we've reached the limit, none are left.
>>> This patch uses the CLONE_DETACHED flag (unused) as a marker to extend the
>> Are we pretty sure that there is no legacy software out there which has
>> continued to specify CLONE_DETACHED since the kernel ignores it?
>
> Please see -mm's
> clone-prepare-to-recycle-clone_detached-and-clone_stopped.patch
>
> That patch has been cut back to only recycle CLONE_STOPPED because there
> indeed was software out there which is setting CLONE_DETACHED.
>
> See http://linux.derkeiler.com/Mailing-Lists/Kernel/2007-11/msg04293.html
bummer, I used the wrong one :/
Thanks,
C.