This comment has been around since c/s 1372bca0615 in 2004.  It is stale, as
it predates the introduction of struct vcpu.

It is not obvious that it was even correct at the time.  Where a vcpu (domain
at the time) has been configured to run is unrelated to construct the domain's
initial pagetables, etc.

Signed-off-by: Andrew Cooper <andrew.cooper3@citrix.com>
---
CC: Jan Beulich <JBeulich@suse.com>
CC: Roger Pau Monné <roger.pau@citrix.com>
CC: Wei Liu <wl@xen.org>

Almost...  I'm not entirely sure NUMA memory allocation is plumbed through
correctly, but even that still has nothing to do with v->processor
---
 xen/arch/x86/pv/dom0_build.c | 1 -
 1 file changed, 1 deletion(-)

diff --git a/xen/arch/x86/pv/dom0_build.c b/xen/arch/x86/pv/dom0_build.c
index d79503d6a9..f7165309a2 100644
--- a/xen/arch/x86/pv/dom0_build.c
+++ b/xen/arch/x86/pv/dom0_build.c
@@ -616,7 +616,6 @@ int __init dom0_construct_pv(struct domain *d,
         v->arch.pv.event_callback_cs    = FLAT_COMPAT_KERNEL_CS;
     }
 
-    /* WARNING: The new domain must have its 'processor' field filled in! */
     if ( !is_pv_32bit_domain(d) )
     {
         maddr_to_page(mpt_alloc)->u.inuse.type_info = PGT_l4_page_table;
-- 
2.11.0

On 29.10.2020 15:00, Andrew Cooper wrote:
> This comment has been around since c/s 1372bca0615 in 2004.  It is stale, as
> it predates the introduction of struct vcpu.

That commit only moved it around; it's 22a857bde9b8 afaics from
early 2003 where it first appeared, where it had a reason:

    /*
     * WARNING: The new domain must have its 'processor' field
     * filled in by now !!
     */
    phys_l2tab = ALLOC_FRAME_FROM_DOMAIN();
    l2tab = map_domain_mem(phys_l2tab);
    memcpy(l2tab, idle_pg_table[p->processor], PAGE_SIZE);

But yes, the comment has been stale for a long time, and I've
been wondering a number of times what it was supposed to tell
me. (I think it was already stale at the point the comment
first got altered, in 3072fef54df8.)

> It is not obvious that it was even correct at the time.  Where a vcpu (domain
> at the time) has been configured to run is unrelated to construct the domain's
> initial pagetables, etc.
> 
> Signed-off-by: Andrew Cooper <andrew.cooper3@citrix.com>

Acked-by: Jan Beulich <jbeulich@suse.com>

Jan

On 29/10/2020 14:37, Jan Beulich wrote:
> On 29.10.2020 15:00, Andrew Cooper wrote:
>> This comment has been around since c/s 1372bca0615 in 2004.  It is stale, as
>> it predates the introduction of struct vcpu.
> That commit only moved it around; it's 22a857bde9b8 afaics from
> early 2003 where it first appeared, where it had a reason:
>
>     /*
>      * WARNING: The new domain must have its 'processor' field
>      * filled in by now !!
>      */
>     phys_l2tab = ALLOC_FRAME_FROM_DOMAIN();
>     l2tab = map_domain_mem(phys_l2tab);
>     memcpy(l2tab, idle_pg_table[p->processor], PAGE_SIZE);

Oh yes - my simple search didn't spot the reformat.

>
> But yes, the comment has been stale for a long time, and I've
> been wondering a number of times what it was supposed to tell
> me. (I think it was already stale at the point the comment
> first got altered, in 3072fef54df8.)

Looks like it became stale with 99db02d5097 "Remove CPU-dependent
page-directory entries." which drops the per-cpu idle_pg_table.

>
>> It is not obvious that it was even correct at the time.  Where a vcpu (domain
>> at the time) has been configured to run is unrelated to construct the domain's
>> initial pagetables, etc.
>>
>> Signed-off-by: Andrew Cooper <andrew.cooper3@citrix.com>
> Acked-by: Jan Beulich <jbeulich@suse.com>

Thanks.  I'll update the commit message.

~Andrew