Xen devel - Mar 2009 - alloc_heap_pages rejecting several order=6 requests early in boot (not tmem)

Due to some debug code I have in tmem (but otherwise
completely unrelated to tmem), I noticed that,
early in dom0 boot, four requests to alloc_heap_pages
for order=6 are getting rejected.  The allocation
requests are coming from do_memory_op which is
reached via a hypercall.  From console output,
the hypercalls are coming after the dom0 kernel does:

(XEN) Freed 132kB init memory

and the dom0 console output immediately following is:

PCI: BIOS: Bug: MCFG area at e0000000 is not E820-reserved
PCI: Not using MMCONFIG
(XEN) io_apic.c2139:

which may or may not be related to the allocation rejections.

Dom0 boot and everything else proceed fine, so apparently
whatever is requesting the memory recovers or uses some
backup approach (e.g. getting smaller chunks of memory?).

I''m just reporting this because the allocation failures
are otherwise silent, and might be indicative of some
bug (one would think that allocating a 256K chunk
of memory during dom0 boot would always succeed...)
or something that might reduce functionality or performance
in dom0.

I''ve been seeing this for some time (weeks at least) but
not sure exactly when it started.  It''s probably easy
to reproduce by flagging failures in alloc_heap_pages
or turning on gdprintk in memory.c, but if not let
me know and I will look further.

Thanks,
Dan

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

On 20/03/2009 16:34, "Dan Magenheimer"
<dan.magenheimer@oracle.com> wrote:
> I''m just reporting this because the allocation failures
> are otherwise silent, and might be indicative of some
> bug (one would think that allocating a 256K chunk
> of memory during dom0 boot would always succeed...)
> or something that might reduce functionality or performance
> in dom0.
These will be swiotlb trying to get the lowest memory it possibly can. It
basically makes allocation calls with increasing acceptable address width
until the request succeeds. So these are harmless I believe.

 Thanks,
 Keir



_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel