Manuel Bouyer wrote:
On Wed, May 21, 2008 at 01:45:13AM +0200, Jean-Yves Migeon wrote:[...]Since I am using xpmap_phys_to_machine_mapping to build it, is it ok to put the "arch.pfn_to_mfn_frame_list_list" initialization in pmap_bootstrap code? Not too early in boot process?No, it's the right place to do it.
So, here is what the code looks like, after some tryouts. See the diff.Just for information; this is not for a commit. Its current place is totaly unsuitable (hypervisor_attach), but as I am using memory allocators, I put it late during boot up process.
When applying it, looks like that the error with xm dump-core disappeared, now I am getting some VCPU error:
current# xm dump-core 1 /root/core Dumping core of domain: 1 ...Error: Failed to dump core: (1, 'Internal error', 'No VCPU context could be grabbed (14 = Bad address)')
Usage: xm dump-core [-L|--live] [-C|--crash] <Domain> [Filename] ...I do not presently understand what such an error means (failed to access vcpu_info[], huh?), I will investigate a bit. Debugging xend to see what raises this error ends in a panic() (see kern/36183).
Question is: is there any other possibility than using memory allocators for my code? xpmap bootstrap happens early during boot up, and I don't think that using malloc/kmem_alloc stuff before mm init is possible.
As always, comments are welcomed :) -- Jean-Yves Migeon jean-yves.migeon%espci.fr@localhost