On 06/19/2017 02:52 PM, David Hildenbrand wrote: > On 19.06.2017 14:47, Christian Borntraeger wrote: >> On 06/19/2017 02:41 PM, David Hildenbrand wrote: >>> On 19.06.2017 14:33, Christian Borntraeger wrote: >>>> On 06/19/2017 02:05 PM, David Hildenbrand wrote: >>>>> On 19.06.2017 12:03, David Hildenbrand wrote: >>>>>> On 15.06.2017 22:37, Richard Henderson wrote: >>>>>>> There are no uses in a Linux system with which to test, >>>>>>> but it Looks Right by my reading of the PoO. >>>>>> >>>>>> I am using next.git/master with this patch applied: >>>>>> https://git.kernel.org/pub/scm/linux/kernel/git/s390/linux.git/commit/?h=features&id=8aa8680aa383bf6e2ac >>>>>> >>>>>> I am using QEMU with the mvcos patch and your patch applied (and a patch >>>>>> that allows enabling csst/csst2). >>>>>> >>>>>> I am using the following qemu command line: >>>>>> >>>>>> #!/bin/bash >>>>>> /home/dhildenb/git/qemu/s390x-softmmu/qemu-system-s390x \ >>>>>> -nographic -nodefaults -machine s390-ccw-virtio,accel=tcg \ >>>>>> -cpu qemu,mvcos=on,stfle=on,ldisp=on,ldisphp=on,\ >>>>>> eimm=on,stckf=on,csst=on,csst2=on,ginste=on,exrl=on\ >>>>>> -m 256M -smp 1 -chardev stdio,id=con0 \ >>>>>> -device sclpconsole,chardev=con0 \ >>>>>> -kernel vmlinux -initrd /home/dhildenb/initrd.debian >>>>>> >>>>>> Right now, I can start a z9 compiled kernel. >>>>>> >>>>>> When trying to start a z10 compiled kernel (which generates many csst), >>>> >>>> >>>> I would be very surprised if the kernel would contain any csst. gcc does >>>> not >>>> emit csst and the kernel source also does not contain it. >>>> >>> >>> I only did a grep on the objdump output: >>> >>> t460s: ~/git/linux-s390 next $ /usr/bin/s390x-linux-gnu-objdump -D >>> vmlinux | grep csst >>> 912826: c8 e2 f1 7c 56 02 csst 380(%r15),1538(%r5),%r14 >>> 954684: c8 62 dc 35 2b 65 csst 3125(%r13),2917(%r2),%r6 >>> 95e6e4: c8 a2 1c 76 0a 60 csst 3190(%r1),2656,%r10 >>> 95f68a: c8 b2 c9 b3 3d 47 csst 2483(%r12),3399(%r3),%r11 >>> 96067a: c8 42 d3 59 da 50 csst 857(%r13),2640(%r13),%r4 >>> 963642: c8 72 73 c9 1a a0 csst 969(%r7),2720(%r1),%r7 >>> 9656de: c8 12 d3 09 7a a0 csst 777(%r13),2720(%r7),%r1 >>> 9676a6: c8 32 6d 97 84 7e csst 3479(%r6),1150(%r8),%r3 >>> 9d470a: c8 a2 70 11 74 02 csst 17(%r7),1026(%r7),%r10 >>> 9d6c4a: c8 a2 de 0c 54 4a csst 3596(%r13),1098(%r5),%r10 >>> 9e3af8: c8 a2 de 09 54 73 csst 3593(%r13),1139(%r5),%r10 >>> 9e3b02: c8 a2 de 0f 54 73 csst 3599(%r13),1139(%r5),%r10 >>> 9e7992: c8 a2 de 0c 54 d4 csst 3596(%r13),1236(%r5),%r10 >>> 9e79ea: c8 a2 40 f6 c3 0e csst 246(%r4),782(%r12),%r10 >>> 9e7e3c: c8 a2 40 6c d1 74 csst 108(%r4),372(%r13),%r10 >>> 9e8036: c8 a2 de 0d 54 cd csst 3597(%r13),1229(%r5),%r10 >>> 9e81ea: c8 a2 40 63 2f b8 csst 99(%r4),4024(%r2),%r10 >>> 9e81fe: c8 a2 de 0f 54 68 csst 3599(%r13),1128(%r5),%r10 >>> 9e8e10: c8 72 93 83 69 bd csst 899(%r9),2493(%r6),%r7 >>> 9e8ea4: c8 72 c6 04 54 63 csst 1540(%r12),1123(%r5),%r7 >>> 9e8eae: c8 72 c6 f1 98 77 csst 1777(%r12),2167(%r9),%r7 >>> 9e8ebc: c8 72 93 ba f5 07 csst 954(%r9),1287(%r15),%r7 >>> a0702e: c8 a2 14 74 6e 5f csst 1140(%r1),3679(%r6),%r10 >>> a083ea: c8 a2 73 08 74 da csst 776(%r7),1242(%r7),%r10 >>> a0ec06: c8 a2 f8 f6 c3 08 csst 2294(%r15),776(%r12),%r10 >>> a11890: c8 a2 f8 fa 5f ac csst 2298(%r15),4012(%r5),%r10 >>> a11b6e: c8 a2 73 0a 74 74 csst 778(%r7),1140(%r7),%r10 >>> a11be0: c8 a2 f8 fa 5f ac csst 2298(%r15),4012(%r5),%r10 >>> a11ef4: c8 a2 73 0b b3 7c csst 779(%r7),892(%r11),%r10 >>> [...] >>> 14c9e5a: c8 02 d0 2d 00 0d csst 45(%r13),13,%r0 >>> >>> >>> That made me assume we have csst in the kernel :) >> >> you used -D (which also disassembles data). Do you still see any csst with >> -d ? >> > > ... probably I should have lunch now :) > > No csst, therefore the panic I am seeing is unrelated to this ... > (I wonder why CSST is a required kernel facility if nobody uses it? hm)
I guess because in theory gcc could use that instruction for -march=z9-ec. It does not seem to be hypervisor-managed and should be available on all z9 guests.
