Yes. Placing APs in one hlt-loop is better in one ExitBootServices callback function registered by CpuDxe driver.
In long mode, page table still be required to run hlt-loop instruction. To avoid page table (in Boot Service Data range also) crashed, we could place APs hlt in protected mode. Jeff -----Original Message----- From: Justen, Jordan L Sent: Friday, November 14, 2014 3:02 PM To: Fan, Jeff; edk2-devel@lists.sourceforge.net Subject: RE: [edk2] qemu + osx + smp vs. dee9376ffc6e5522bb5e50512b521952a5421fa5 On 2014-11-13 22:10:42, Fan, Jeff wrote: > Gabriel, > > After APs initialized, APs will run one loop in > ProcessorToIdleState() to check the if BSP assigned task to them. > > After ExitBootService() invoked, the boot services data will not be > safe and the AP running code maybe corrupted. > > Could you check if ExitBootService() is invoked or not when issue > happened? Hmm, I notice CpuDxe doesn't have an ExitBootServices callback yet. If 1 or more APs are found we should consider having the BSP put the APs back into the wait-for-sipi state at ExitBootServices. Otherwise, these will continue to loop using boot services code and data memory. -Jordan > -----Original Message----- > From: Gabriel Somlo [mailto:gso...@gmail.com] > Sent: Friday, November 14, 2014 1:15 PM > To: Justen, Jordan L > Cc: edk2-devel@lists.sourceforge.net > Subject: [edk2] qemu + osx + smp vs. > dee9376ffc6e5522bb5e50512b521952a5421fa5 > > Jordan, > > Commit dee9376ffc6e5522bb5e50512b521952a5421fa5 appears to break osx smp mode > (single-cpu on Mavericks seems to still work fine). > > Fedora 20 and 21 live still work fine in either single- or multi-cpu > configurations. > > I realize this must be an annoying bug report, since starting a qemu osx > guest requires not-yet-committed patches (e.g. Reza's pending ahci series, > and another bunch of osx specific stuff)... > > For the record, my command line is: > > bin/qemu-system-x86_64 -enable-kvm -m 2048 -cpu core2duo -smp 4,cores=2 \ > -machine q35 -device ide-drive,bus=ide.2,drive=MacHDD \ > -drive id=MacHDD,if=none,snapshot=on,file=./mac_10.9.img \ > -device isa-applesmc,osk="secret-smc-osk-string" \ > -netdev user,id=usr0 -device e1000-82545em,netdev=usr0,id=vnet0 \ > -usb -device usb-kbd -device usb-mouse \ > -bios OVMF.fd \ > -smbios type=0,vendor="ovmf",version="svn.16189",date="09/29/14",uefi=on \ > -debugcon file:debug.log -global isa-debugcon.iobase=0x402 \ > -monitor stdio > > And it hangs right before switching to graphics mode (so probably during > apple's boot.efi) with the following message: > > > efiboot loaded from device: > Acpi(PNP0A03,0)/Pci(1F|2)/SATA(2,0)/HD(Part2,SigF402FAFD-C5D8-49C3-953 > C-91354746C321) > boot file path: > .Loading kernel cache file > 'System\Library\Caches\com.apple.kext.caches\Startup\kernelcache'..... > root device uuid is 'D3720931-EC86-377F-8606-DB54F76A1050' > > > To boot with PIIX instead of q35, simply > > s/-machine q35/-device ahci,id=ide/ > > in the command line above. Then, it reboots after logging a similar message. > > > The patch in question is a one-liner adding a "StartApsStackless()" > call to UefiCpuPkg/CpuDxe/CpuMp.c, and osx obviously dislikes > *something* about the state of the APs when it finds them, but beyond that I > have no idea what might be wrong. > > > Jordan, if by any chance you have a hunch and would like me to test an > alternative way of doing things for qemu/osx guest compatibility, I'd be > happy to. > > > Otherwise, this should serve as a heads-up for Reza: now that ovmf q35 > support is upstream, once your ahci patches are applied, this will be > something you'll end up running into when submitting osx patches next > :) > > Thanks again, > --Gabriel > > ---------------------------------------------------------------------- > -------- Comprehensive Server Monitoring with Site24x7. > Monitor 10 servers for $9/Month. > Get alerted through email, SMS, voice calls or mobile push notifications. > Take corrective actions from your mobile device. > http://pubads.g.doubleclick.net/gampad/clk?id=154624111&iu=/4140/ostg. > clktrk _______________________________________________ > edk2-devel mailing list > edk2-devel@lists.sourceforge.net > https://lists.sourceforge.net/lists/listinfo/edk2-devel ------------------------------------------------------------------------------ Comprehensive Server Monitoring with Site24x7. Monitor 10 servers for $9/Month. Get alerted through email, SMS, voice calls or mobile push notifications. Take corrective actions from your mobile device. http://pubads.g.doubleclick.net/gampad/clk?id=154624111&iu=/4140/ostg.clktrk _______________________________________________ edk2-devel mailing list edk2-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/edk2-devel
