On 10/31/15 8:06 , 龙白滔 wrote: > Attached two pictures show the screen shot when the system was suspend during > initialization. The rest pictures show more information about the CPU from > BIOS or CPU back. > > > Do you guys have similar experience in trying OEMs CPUs? We have around 1000 > FACEBOOK OCP servers and want to run all of them on SmartOS.
Hi, We don't have any particular experience with this platform at Joyent. I suspect that there's nothing specific to the CPU in question, but rather something else in the platform is tripping up the initialization process. There are a couple of things I'd suggest. First, does this platform have IPMI capabilities or is there someway to inject an NMI? If so, then what I'd do is boot into kmdb, continue with :c, and then once one of these servers hang, inject an NMI and see where it ends up. If you don't have that, I'd consider single doing a bit of a binary single step. Keep in mind, you did a step over of the call to main. In theory that function never returns from the kernel, so you're not going to get very far that way. Another trick I sometimes use is to set moddebug to 0x80000001 which will cause us to hit a breakpoint after every module is loaded, thus giving you a chance to see where the last one comes into play. If you can get a bit more information about where you're hanging then we can help try and fix what's going on. Robert ------------------------------------------- smartos-discuss Archives: https://www.listbox.com/member/archive/184463/=now RSS Feed: https://www.listbox.com/member/archive/rss/184463/25769125-55cfbc00 Modify Your Subscription: https://www.listbox.com/member/?member_id=25769125&id_secret=25769125-7688e9fb Powered by Listbox: http://www.listbox.com
