Re: [PATCH] PPC64 mmu_context_init needs to run earlier

[Nathan Lynch <nathanl@austin.ibm.com>]

On Tue, 2004-11-02 at 00:13, Andrew Morton wrote:
> Paul Mackerras <paulus@samba.org> wrote:
> >
> > This patch changes mmu_context_init to be called as a core_initcall
> >  rather than an arch_initcall, since mmu_context_init needs to run
> >  before we try to run any userspace processes, and arch_initcall was
> >  found to be too late.
> 
> Here we go again...
> 
> I don't see why your patch fixes the problem.  do_basic_setup() calls
> driver_init() prior to any initcalls being run, and driver_init() can call
> /sbin/hotplug.

My original submission was more of a "works for me, not sure it's
correct" sort of thing.  Sorry I wasn't clear.

How about this instead:


Using idr_get_new_above in init_new_context lets us get rid of an
awkward init function which wasn't running early enough in boot
anyway.

Signed-off-by: Nathan Lynch <nathanl@austin.ibm.com>


---


diff -puN arch/ppc64/mm/init.c~ppc64-mmu-context-use-idr_get_new_above arch/ppc64/mm/init.c
--- linux-2.6.10-rc1-bk12/arch/ppc64/mm/init.c~ppc64-mmu-context-use-idr_get_new_above	2004-11-02 14:22:23.000000000 -0600
+++ linux-2.6.10-rc1-bk12-nathanl/arch/ppc64/mm/init.c	2004-11-02 14:29:09.000000000 -0600
@@ -489,7 +489,7 @@ again:
 		return -ENOMEM;
 
 	spin_lock(&mmu_context_lock);
-	err = idr_get_new(&mmu_context_idr, NULL, &index);
+	err = idr_get_new_above(&mmu_context_idr, NULL, 0, &index);
 	spin_unlock(&mmu_context_lock);
 
 	if (err == -EAGAIN)
@@ -518,19 +518,6 @@ void destroy_context(struct mm_struct *m
 	hugetlb_mm_free_pgd(mm);
 }
 
-static int __init mmu_context_init(void)
-{
-	int index;
-
-	/* Reserve the first (invalid) context*/
-	idr_pre_get(&mmu_context_idr, GFP_KERNEL);
-	idr_get_new(&mmu_context_idr, NULL, &index);
-	BUG_ON(0 != index);
-
-	return 0;
-}
-arch_initcall(mmu_context_init);
-
 /*
  * Do very early mm setup.
  */

_