Re: git: e8b2980e4a12 - main - i386 doreti: stop saving/restoring %ecx around calls into C
Date: Fri, 02 Sep 2022 14:48:22 UTC
On Fri, 2 Sep 2022 17:05:27 +0300 Konstantin Belousov <kostikbel@gmail.com> wrote: > On Fri, Sep 02, 2022 at 03:02:31PM +0200, Tijl Coosemans wrote: >> On Wed, 24 Aug 2022 19:25:12 GMT Konstantin Belousov <kib@FreeBSD.org> >> wrote: >>> The branch main has been updated by kib: >>> >>> URL: https://cgit.FreeBSD.org/src/commit/?id=e8b2980e4a126407e15d33ee8d987385c8321261 >>> >>> commit e8b2980e4a126407e15d33ee8d987385c8321261 >>> Author: Konstantin Belousov <kib@FreeBSD.org> >>> AuthorDate: 2022-08-18 04:07:29 +0000 >>> Commit: Konstantin Belousov <kib@FreeBSD.org> >>> CommitDate: 2022-08-24 19:12:05 +0000 >>> >>> i386 doreti: stop saving/restoring %ecx around calls into C >>> >>> There is no reason to do this. Instead just calculate it later. >>> >>> Reviewed by: jhb >>> Tested by: pho >>> Sponsored by: The FreeBSD Foundation >>> MFC after: 1 week >>> Differential revision: https://reviews.freebsd.org/D36302 >>> --- >>> sys/i386/i386/exception.s | 17 ++++++++--------- >>> 1 file changed, 8 insertions(+), 9 deletions(-) >>> >>> diff --git a/sys/i386/i386/exception.s b/sys/i386/i386/exception.s >>> index 5eef9c1d512c..8279a4602ebf 100644 >>> --- a/sys/i386/i386/exception.s >>> +++ b/sys/i386/i386/exception.s >>> @@ -518,22 +518,21 @@ doreti_exit: >>> je doreti_iret_nmi >>> cmpl $T_TRCTRAP, TF_TRAPNO(%esp) >>> je doreti_iret_nmi >>> - movl $TF_SZ, %ecx >>> testl $PSL_VM,TF_EFLAGS(%esp) >>> - jz 1f /* PCB_VM86CALL is not set */ >>> - addl $VM86_STACK_SPACE, %ecx >>> - jmp 2f >>> -1: testl $SEL_RPL_MASK, TF_CS(%esp) >>> + jnz 1f /* PCB_VM86CALL is not set */ >>> + testl $SEL_RPL_MASK, TF_CS(%esp) >>> jz doreti_popl_fs >>> -2: movl $handle_ibrs_exit,%eax >>> - pushl %ecx /* preserve enough call-used regs */ >>> +1: movl $handle_ibrs_exit,%eax >>> call *%eax >>> movl mds_handler,%eax >>> call *%eax >>> - popl %ecx >>> movl %esp, %esi >>> movl PCPU(TRAMPSTK), %edx >>> - subl %ecx, %edx >>> + movl $TF_SZ, %ecx >>> + testl $PSL_VM,TF_EFLAGS(%esp) >>> + jz 2f /* PCB_VM86CALL is not set */ >>> + addl $VM86_STACK_SPACE, %ecx >>> +2: subl $TF_SZ, %edx >> >> Shouldn't this be "subl %ecx, %edx"? > > You are right, please fix. Fixed in cfdc649e455b.