x86: print MCAs to the console before MCA initialization

[jtl]

Summary

This closes a theoretical race if a message arrives after the interrupts are configured, but before the MCA system is fully initialized.

Note that I am not 100% sure this can actually happen, since I am not 100% sure we will ever enable interrupts before we run mca_setup(). Some other hints in the code (for example, the previous checks for 'cold') make it seem possible.

[markj]

Note that I am not 100% sure this can actually happen, since I am not 100% sure we will ever enable interrupts before we run mca_setup().

I don't quite follow, MCEs will be delivered even if interrupts are disabled. _mca_init() enables delivery of MCEs at SI_SUB_CPU time; what's wrong with calling mca_record_entry() after that point?

[jtl]

In D52943#1209843, @markj wrote:

Note that I am not 100% sure this can actually happen, since I am not 100% sure we will ever enable interrupts before we run mca_setup().

I don't quite follow, MCEs will be delivered even if interrupts are disabled. _mca_init() enables delivery of MCEs at SI_SUB_CPU time; what's wrong with calling mca_record_entry() after that point?

Doh. You are correct. The checks for cold (now mca_startup_done) were only for task queues and not other aspects of this.