tree: https://git.kernel.org/pub/scm/linux/kernel/git/luto/linux.git x86/entry head: 69982f2c73a199503541565a5533583ef07adcf7 commit: bad87f44f688bdddcb377bc4434c90d1c01e945e [5/10] x86/entry: Use the high bits of regs->cs to store the entry type config: i386-allyesconfig (attached as .config) compiler: gcc-9 (Debian 9.3.0-13) 9.3.0 reproduce (this is a W=1 build): git checkout bad87f44f688bdddcb377bc4434c90d1c01e945e # save the attached .config to linux build tree make W=1 ARCH=i386 If you fix the issue, kindly add following tag as appropriate Reported-by: kernel test robot <lkp(a)intel.com&gt; All errors (new ones prefixed by >>, old ones prefixed by <<): arch/x86/entry/entry_32.S: Assembler messages: vim +930 arch/x86/entry/entry_32.S 876 877 /* 878 * 32-bit SYSENTER entry. 879 * 880 * 32-bit system calls through the vDSO's __kernel_vsyscall enter here 881 * if X86_FEATURE_SEP is available. This is the preferred system call 882 * entry on 32-bit systems. 883 * 884 * The SYSENTER instruction, in principle, should *only* occur in the 885 * vDSO. In practice, a small number of Android devices were shipped 886 * with a copy of Bionic that inlined a SYSENTER instruction. This 887 * never happened in any of Google's Bionic versions -- it only happened 888 * in a narrow range of Intel-provided versions. 889 * 890 * SYSENTER loads SS, ESP, CS, and EIP from previously programmed MSRs. 891 * IF and VM in RFLAGS are cleared (IOW: interrupts are off). 892 * SYSENTER does not save anything on the stack, 893 * and does not save old EIP (!!!), ESP, or EFLAGS. 894 * 895 * To avoid losing track of EFLAGS.VM (and thus potentially corrupting 896 * user and/or vm86 state), we explicitly disable the SYSENTER 897 * instruction in vm86 mode by reprogramming the MSRs. 898 * 899 * Arguments: 900 * eax system call number 901 * ebx arg1 902 * ecx arg2 903 * edx arg3 904 * esi arg4 905 * edi arg5 906 * ebp user stack 907 * 0(%ebp) arg6 908 */ 909 SYM_FUNC_START(entry_SYSENTER_32) 910 /* 911 * On entry-stack with all userspace-regs live - save and 912 * restore eflags and %eax to use it as scratch-reg for the cr3 913 * switch. 914 */ 915 pushfl 916 pushl %eax 917 BUG_IF_WRONG_CR3 no_user_check=1 918 SWITCH_TO_KERNEL_CR3 scratch_reg=%eax 919 popl %eax 920 popfl 921 922 /* Stack empty again, switch to task stack */ 923 movl TSS_entry2task_stack(%esp), %esp 924 925 .Lsysenter_past_esp: 926 pushl $__USER_DS /* pt_regs->ss */ 927 pushl %ebp /* pt_regs->sp (stashed in bp) */ 928 pushfl /* pt_regs->flags (except IF = 0) */ 929 orl $X86_EFLAGS_IF, (%esp) /* Fix IF */ 931 pushl $0 /* pt_regs->ip = 0 (placeholder) */ 932 pushl %eax /* pt_regs->orig_ax */ 933 SAVE_ALL pt_regs_ax=$-ENOSYS /* save rest, stack already switched */ 934 935 /* 936 * SYSENTER doesn't filter flags, so we need to clear NT, AC 937 * and TF ourselves. To save a few cycles, we can check whether 938 * either was set instead of doing an unconditional popfq. 939 * This needs to happen before enabling interrupts so that 940 * we don't get preempted with NT set. 941 * 942 * If TF is set, we will single-step all the way to here -- do_debug 943 * will ignore all the traps. (Yes, this is slow, but so is 944 * single-stepping in general. This allows us to avoid having 945 * a more complicated code to handle the case where a user program 946 * forces us to single-step through the SYSENTER entry code.) 947 * 948 * NB.: .Lsysenter_fix_flags is a label with the code under it moved 949 * out-of-line as an optimization: NT is unlikely to be set in the 950 * majority of the cases and instead of polluting the I$ unnecessarily, 951 * we're keeping that code behind a branch which will predict as 952 * not-taken and therefore its instructions won't be fetched. 953 */ 954 testl $X86_EFLAGS_NT|X86_EFLAGS_AC|X86_EFLAGS_TF, PT_EFLAGS(%esp) 955 jnz .Lsysenter_fix_flags 956 .Lsysenter_flags_fixed: 957 958 movl %esp, %eax 959 call do_fast_syscall_32 960 /* XEN PV guests always use IRET path */ 961 ALTERNATIVE "testl %eax, %eax; jz .Lsyscall_32_done", \ 962 "jmp .Lsyscall_32_done", X86_FEATURE_XENPV 963 964 STACKLEAK_ERASE 965 966 /* Opportunistic SYSEXIT */ 967 968 /* 969 * Setup entry stack - we keep the pointer in %eax and do the 970 * switch after almost all user-state is restored. 971 */ 972 973 /* Load entry stack pointer and allocate frame for eflags/eax */ 974 movl PER_CPU_VAR(cpu_tss_rw + TSS_sp0), %eax 975 subl $(2*4), %eax 976 977 /* Copy eflags and eax to entry stack */ 978 movl PT_EFLAGS(%esp), %edi 979 movl PT_EAX(%esp), %esi 980 movl %edi, (%eax) 981 movl %esi, 4(%eax) 982 983 /* Restore user registers and segments */ 984 movl PT_EIP(%esp), %edx /* pt_regs->ip */ 985 movl PT_OLDESP(%esp), %ecx /* pt_regs->sp */ 986 1: mov PT_FS(%esp), %fs 987 PTGS_TO_GS 988 989 popl %ebx /* pt_regs->bx */ 990 addl $2*4, %esp /* skip pt_regs->cx and pt_regs->dx */ 991 popl %esi /* pt_regs->si */ 992 popl %edi /* pt_regs->di */ 993 popl %ebp /* pt_regs->bp */ 994 995 /* Switch to entry stack */ 996 movl %eax, %esp 997 998 /* Now ready to switch the cr3 */ 999 SWITCH_TO_USER_CR3 scratch_reg=%eax 1000 1001 /* 1002 * Restore all flags except IF. (We restore IF separately because 1003 * STI gives a one-instruction window in which we won't be interrupted, 1004 * whereas POPF does not.) 1005 */ 1006 btrl $X86_EFLAGS_IF_BIT, (%esp) 1007 BUG_IF_WRONG_CR3 no_user_check=1 1008 popfl 1009 popl %eax 1010 1011 /* 1012 * Return back to the vDSO, which will pop ecx and edx. 1013 * Don't bother with DS and ES (they already contain __USER_DS). 1014 */ 1015 sti 1016 sysexit 1017 --- 0-DAY CI Kernel Test Service, Intel Corporation https://lists.01.org/hyperkitty/list/kbuild-all@lists.01.org