While this changeset appears to finally address the cpu_possible_map problems
(according to other reports on this list)
I wonder if this isn''t incomplete. With prefill_possible_map() now
being called from setup_arch() (why can''t it be
called at the same point as native calls it when CONFIG_HOTPLUG_CPU?), the
immediately following call to
smp_prepare_boot_cpu() (from start_kernel) will destroy this information again.
With the effort of not setting up
per-CPU information for impossible CPUs (some of this exists in our internal
trees, not sure how much has been posted to
mainline) this would be prone to break again soon.
Also, some of other code seems then superfluous or even ill:
- drivers/xen/core/smpboot.c: smp_prepare_cpus() should not initialize
cpu_possible_map again, but should instead have
its main loop controlled by this bit vector (and VCPUOP_is_up doesn''t
need to be called then anymore, too)
- arch/{i386,x86_64}/kernel/mpparse-xen.c: the setting of num_processors was and
is completely out of sync with the
real set of VCPUs used; I even wonder what purpose the whole file still serves
Jan
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel
On 8 Feb 2006, at 09:37, Jan Beulich wrote:> Also, some of other code seems then superfluous or even ill: > - drivers/xen/core/smpboot.c: smp_prepare_cpus() should not initialize > cpu_possible_map again, but should instead have > its main loop controlled by this bit vector (and VCPUOP_is_up doesn''t > need to be called then anymore, too)I now fixed smp_prepare_boot_cpu/smp_prepare_cpus to respect the cpu_possible_map.> - arch/{i386,x86_64}/kernel/mpparse-xen.c: the setting of > num_processors was and is completely out of sync with the > real set of VCPUs used; I even wonder what purpose the whole file > still servesYes, we would like to get rid of at least the CPU/local-APIC portions of mpparse.c. If there''s no IRQ/IO_APIC info in those tables we can get rid of the whole file. -- Keir _______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel