Hi Anthony. Some comments. Please see the below
On Thu, May 08, 2008 at 04:51:56PM +0800, Xu, Anthony wrote:
> diff -r f2457c7aff8d xen/arch/ia64/vmx/optvfault.S
> --- a/xen/arch/ia64/vmx/optvfault.S Fri Apr 25 20:13:52 2008 +0900
> +++ b/xen/arch/ia64/vmx/optvfault.S Thu May 08 16:23:42 2008 +0800
> @@ -20,6 +20,135 @@
> #include <asm/virt_event.h>
> #include <asm-ia64/vmx_mm_def.h>
> #include <asm-ia64/vmx_phy_mode.h>
> +#include "entry.h"
> +
> +// r21 current
> +// r23 : b0
> +// r31 : pr
> +
> +#define VMX_VIRT_SAVE \
> + mov r27=ar.rsc; /* M */
> \
> + ;;
>
> \
> + cover; /* B;; (or nothing) */
> \
> + ;;
> \
> + /* switch from user to kernel RBS: */
> \
> + invala; /* M */
> \
> + ;;
> \
> + mov ar.rsc=0; /* set enforced lazy mode, pl 0, little-endian,
> loadrs=0 */ \
> + ;;
> \
> + mov.m r26=ar.rnat;
> \
> + movl r28=IA64_RBS_OFFSET; /* compute base of RBS */
> \
> + ;;
> \
> + mov r22=ar.bspstore; /* save ar.bspstore */
> \
> + add r28=r28,r21;
> \
> + ;;
> \
> + mov ar.bspstore=r28; /* switch to kernel RBS
> */ \
> + ;;
> \
> + mov r18=ar.bsp;
> \
> + mov ar.rsc=0x3; /* set eager mode, pl 0, little-endian, loadrs=0
> */ \
> + ;;
> \
> + alloc r32=ar.pfs,24,0,3,0 /* save pfs */
> \
> + ;;
> \
> + sub r18=r18,r28; /* r18=RSE.ndirty*8 */
> \
> + ;;
> \
> + shl r33=r18,16; /* save loadrs, ompute ar.rsc to be used for
> "loadrs" */ \
> + mov r35=b6; /* save b6 */
> \
> + mov r36=b7; /* save b7 */
> \
> + mov r37=ar.csd; /* save ar.csd */
> \
> + mov r38=ar.ssd; /* save ar.ssd */
> \
> + mov r39=r8; /* save r8 */ \
> + mov r40=r9; /* save r9 */ \
> + mov r41=r10; /* save r10 */ \
> + mov r42=r11; /* save r11 */ \
> + mov r43=r27; /* save ar.rsc */ \
> + mov r44=r26; /* save ar.rnat */ \
> + mov r45=r22; /* save ar.bspstore */ \
> + mov r46=r31; /* save pr */ \
> + mov r47=r23; /* save b0 */ \
> + mov r48=r1; /* save r1 */ \
> + mov r49=r12; /* save r12 */ \
> + mov r50=r13; /* save r13 */ \
> + mov r51=r15; /* save r15 */ \
> + mov r52=r14; /* save r14 */ \
> + mov r53=r2; /* save r2 */
> \
> + mov r54=r3; /* save r3 */ \
> + mov r34=ar.ccv; /* save ar.ccv */ \
> + ;; \
> + movl r1=__gp;
> \
> + movl r29=IA64_STK_OFFSET-IA64_PT_REGS_SIZE-16;
> \
> + ;; \
> + add r12=r29,r21; /* compute base of memory stack */ \
> + mov r13=r21;
> \
> + ;; \
> +{ .mii; /* call vps sync read */ \
> + add r25=IA64_VPD_BASE_OFFSET, r21; \
> + nop 0x0;
> \
> + mov r24=ip; \
> + ;; \
> +}; \
> +{ .mmb; \
> + add r24 = 0x20, r24; \
> + ld8 r25=[r25]; /* read vpd base */ \
> + br.cond.sptk vmx_vps_sync_read; /* call the service */ \
> + ;; \
> +};
> +
> +
Please fix white space before the backslash.
In vmx_optfaults.S 8 tab is consistently used.
> +GLOBAL_ENTRY(ia64_leave_hypervisor_virt)
> + PT_REGS_UNWIND_INFO(0)
Is this correct? I'm just asking just to make it sure.
To be honest I have to dig into the related specifications.
But the register save/restore convention of VMX_VIRT_SAVE() is
different from the one of DO_SAVE_MIN(). So some twist to
the stack unwinder is necessary, isn't it?
> + ;;
> + invala /* M */
> + ;;
> + mov r21 = r13 /* get current */
> + mov b6 = r35 /* restore b6 */
> + mov b7 = r36 /* restore b7 */
> + mov ar.csd = r37 /* restore ar.csd */
> + mov ar.ssd = r38 /* restore ar.ssd */
> + mov r8 = r39 /* restore r8 */
> + mov r9 = r40 /* restore r9 */
> + mov r10 = r41 /* restore r10 */
> + mov r11 = r42 /* restore r11 */
> + mov ar.pfs = r32 /* restore ar.pfs */
> + mov r27 = r43 /* restore ar.rsc */
> + mov r26 = r44 /* restore ar.rnat */
> + mov r25 = r45 /* restore ar.bspstore */
> + mov r23 = r46 /* restore predicates */
> + mov r22 = r47 /* restore b0 */
> + mov r1 = r48 /* restore r1 */
> + mov r12 = r49 /* restore r12 */
> + mov r13 = r50 /* restore r13 */
> + mov r15 = r51 /* restore r15 */
> + mov r14 = r52 /* restore r14 */
> + mov r2 = r53 /* restore r2 */
> + mov r3 = r54 /* restore r3 */
> + mov ar.ccv = r34 /* restore ar.ccv */
> + mov ar.rsc = r33 // load ar.rsc to be used for "loadrs"
please fix white space before those comment and white spaces
around '=' for consistency.
> + ;;
> + alloc r16=ar.pfs,0,0,0,0 // drop current register frame
> + ;;
> + loadrs
> + ;;
> + mov ar.bspstore=r25
> + ;;
> + mov ar.rnat=r26
> + ;;
> + mov ar.rsc=r27
> + adds r18=IA64_VPD_BASE_OFFSET,r21
> + ;;
> + ld8 r25=[r18] //vpd
> + mov r17=r0
> + ;;
> +//vsa_sync_write_start
> + ;;
> + movl r24=ia64_leave_hypervisor_virt_1 // calculate return address
> + br.cond.sptk vmx_vps_sync_write // call the service
> + ;;
> +ia64_leave_hypervisor_virt_1:
> + mov r24=r22
> + mov r31=r23
> + br.cond.sptk vmx_resume_to_guest
> +END(ia64_leave_hypervisor_virt)
> +
>
> #define ACCE_MOV_FROM_AR
> #define ACCE_MOV_FROM_RR
> @@ -30,6 +159,215 @@
> #define ACCE_THASH
>
> // Inputs are: r21 (= current), r24 (= cause), r25 (= insn), r31 (=saved pr)
> +
> +GLOBAL_ENTRY(virtualization_fault_table)
> +{ /* Entry 0 */
> + cmp.eq p2,p0=r0,r0
> + mov b0=r23
> + br.many vmx_virtualization_fault_back
> +}
> +{ /* Entry 1 */
> + cmp.eq p2,p0=r0,r0
> + mov b0=r23
> + br.many vmx_virtualization_fault_back
> +}
> +{ /* Entry 2 */
> + nop.m 0x0
> + mov b0=r23
> + br.many vmx_virtualization_fault_back
> +}
> +{ /* Entry 3 */
> + cmp.eq p2,p0=r0,r0
> + mov b0=r23
> + br.many vmx_asm_mov_from_ar
> +}
> +{ /* Entry 4 */
> + nop.m 0x0
> + mov b0=r23
> + br.many vmx_virtualization_fault_back
> +}
> +{ /* Entry 5 */
> + nop.m 0x0
> + mov b0=r23
> + br.many vmx_virtualization_fault_back
> +}
> +{ /* Entry 6 */
> + cmp.eq p2,p0=r0,r0
> + mov b0=r23
> + br.many vmx_asm_mov_to_psr
> +}
> +{ /* Entry 7 */
> + nop.m 0x0
> + mov b0=r23
> + br.many vmx_virtualization_fault_back
> +}
> +{ /* Entry 8 */
> + nop.m 0x0
> + mov b0=r23
> + br.many vmx_virtualization_fault_back
> +}
> +{ /* Entry 9 */
> + nop.m 0x0
> + mov b0=r23
> + br.many vmx_virtualization_fault_back
> +}
> +{ /* Entry 10 */
> + cmp.eq p2,p0=r0,r0
> + mov b0=r23
> + br.many vmx_asm_mov_to_rr
> +}
> +{ /* Entry 11 */
> + nop.m 0x0
> + mov b0=r23
> + br.many vmx_virtualization_fault_back
> +}
> +{ /* Entry 12 */
> + nop.m 0x0
> + mov b0=r23
> + br.many vmx_virtualization_fault_back
> +}
> +{ /* Entry 13 */
> + nop.m 0x0
> + mov b0=r23
> + br.many vmx_virtualization_fault_back
> +}
> +{ /* Entry 14 */
> + nop.m 0x0
> + mov b0=r23
> + br.many vmx_virtualization_fault_back
> +}
> +{ /* Entry 15 */
> + nop.m 0x0
> + mov b0=r23
> + br.many vmx_virtualization_fault_back
> +}
> +{ /* Entry 16 */
> + nop.m 0x0
> + mov b0=r23
> + br.many vmx_virtualization_fault_back
> +}
> +{ /* Entry 17 */
> + nop.m 0x0
> + mov b0=r23
> + br.many vmx_virtualization_fault_back
> +}
> +{ /* Entry 18 */
> + cmp.eq p2,p0=r0,r0
> + mov b0=r23
> + br.many vmx_asm_mov_from_rr
> +}
> +{ /* Entry 19 */
> + nop.m 0x0
> + mov b0=r23
> + br.many vmx_virtualization_fault_back
> +}
> +{ /* Entry 20 */
> + nop.m 0x0
> + mov b0=r23
> + br.many vmx_virtualization_fault_back
> +}
> +{ /* Entry 21 */
> + nop.m 0x0
> + mov b0=r23
> + br.many vmx_virtualization_fault_back
> +}
> +{ /* Entry 22 */
> + nop.m 0x0
> + mov b0=r23
> + br.many vmx_virtualization_fault_back
> +}
> +{ /* Entry 23 */
> + nop.m 0x0
> + mov b0=r23
> + br.many vmx_virtualization_fault_back
> +}
> +{ /* Entry 24 */
> + cmp.eq p2,p0=r0,r0
> + mov b0=r23
> + br.many vmx_asm_ssm
> +}
> +{ /* Entry 25 */
> + cmp.eq p2,p0=r0,r0
> + mov b0=r23
> + br.many vmx_asm_rsm
> +}
> +{ /* Entry 26 */
> + nop.m 0x0
> + mov b0=r23
> + br.many vmx_virtualization_fault_back
> +}
> +{ /* Entry 27 */
> + nop.m 0x0
> + mov b0=r23
> + br.many vmx_virtualization_fault_back
> +}
> +{ /* Entry 28 */
> + nop.m 0x0
> + mov b0=r23
> + br.many vmx_virtualization_fault_back
> +}
> +{ /* Entry 29 */
> + nop.m 0x0
> + mov b0=r23
> + br.many vmx_virtualization_fault_back
> +}
> +{ /* Entry 30 */
> + nop.m 0x0
> + mov b0=r23
> + br.many vmx_virtualization_fault_back
> +}
> +{ /* Entry 31 */
> + cmp.eq p2,p0=r0,r0
> + mov b0=r23
> + br.many vmx_asm_thash
> +}
> +{ /* Entry 32 */
> + nop.m 0x0
> + mov b0=r23
> + br.many vmx_virtualization_fault_back
> +}
> +{ /* Entry 33 */
> + nop.m 0x0
> + mov b0=r23
> + br.many vmx_virtualization_fault_back
> +}
> +{ /* Entry 34 */
> + nop.m 0x0
> + mov b0=r23
> + br.many vmx_virtualization_fault_back
> +}
> +{ /* Entry 35 */
> + nop.m 0x0
> + mov b0=r23
> + br.many vmx_virtualization_fault_back
> +}
> +{ /* Entry 36 */
> + nop.m 0x0
> + mov b0=r23
> + br.many vmx_virtualization_fault_back
> +}
> +{ /* Entry 37 */
> + cmp.ne p2,p0=r0,r0
> + mov b0=r23
> + br.many vmx_asm_rfi
> +}
> +{ /* Entry 38 */
> + nop.m 0x0
> + mov b0=r23
> + br.many vmx_virtualization_fault_back
> +}
> +{ /* Entry 39 */
> + nop.m 0x0
> + mov b0=r23
> + br.many vmx_virtualization_fault_back
> +}
> +{ /* Entry 40 */
> + nop.m 0x0
> + mov b0=r23
> + br.many vmx_virtualization_fault_back
> +}
> +END(virtualization_fault_table)
> +
Can macro be used instead of repeating of the similar sequence?
>
> ENTRY(vmx_dummy_function)
> br.sptk.many vmx_dummy_function
> @@ -97,12 +435,203 @@ GLOBAL_ENTRY(vmx_vps_resume_handler)
> br.sptk.many b0
> END(vmx_vps_resume_handler)
>
> +//r13 ->vcpu
> +//call with psr.bn = 0
> +GLOBAL_ENTRY(vmx_asm_bsw0)
> + mov r15=ar.unat
> + ;;
> + adds r14=IA64_VPD_BASE_OFFSET,r13
> + ;;
> + ld8 r14=[r14]
> + bsw.1
> + ;;
> + adds r2= IA64_VPD_VB1REG_OFFSET, r14
> + adds r3= IA64_VPD_VB1REG_OFFSET+8, r14
> + ;;
> + .mem.offset 0,0; st8.spill [r2]=r16,16
> + .mem.offset 8,0; st8.spill [r3]=r17,16
> + ;;
> + .mem.offset 0,0; st8.spill [r2]=r18,16
> + .mem.offset 8,0; st8.spill [r3]=r19,16
> + ;;
> + .mem.offset 0,0; st8.spill [r2]=r20,16
> + .mem.offset 8,0; st8.spill [r3]=r21,16
> + ;;
> + .mem.offset 0,0; st8.spill [r2]=r22,16
> + .mem.offset 8,0; st8.spill [r3]=r23,16
> + ;;
> + .mem.offset 0,0; st8.spill [r2]=r24,16
> + .mem.offset 8,0; st8.spill [r3]=r25,16
> + ;;
> + .mem.offset 0,0; st8.spill [r2]=r26,16
> + .mem.offset 8,0; st8.spill [r3]=r27,16
> + ;;
> + .mem.offset 0,0; st8.spill [r2]=r28,16
> + .mem.offset 8,0; st8.spill [r3]=r29,16
> + ;;
> + .mem.offset 0,0; st8.spill [r2]=r30,16
> + .mem.offset 8,0; st8.spill [r3]=r31,16
> + ;;
> + mov r9=ar.unat
> + adds r8=IA64_VPD_VB1NAT_OFFSET, r14
> + ;;
> + st8 [r8]=r9
> + adds r8=IA64_VPD_VB0NAT_OFFSET, r14
> + ;;
> + ld8 r9=[r8]
> + adds r2= IA64_VPD_VB0REG_OFFSET, r14
> + adds r3= IA64_VPD_VB0REG_OFFSET+8, r14
> + ;;
> + mov ar.unat=r9
> + ;;
> + ld8.fill r16=[r2],16
> + ld8.fill r17=[r3],16
> + ;;
> + ld8.fill r18=[r2],16
> + ld8.fill r19=[r3],16
> + ;;
> + ld8.fill r20=[r2],16
> + ld8.fill r21=[r3],16
> + ;;
> + ld8.fill r22=[r2],16
> + ld8.fill r23=[r3],16
> + ;;
> + ld8.fill r24=[r2],16
> + ld8.fill r25=[r3],16
> + ;;
> + ld8.fill r26=[r2],16
> + ld8.fill r27=[r3],16
> + ;;
> + ld8.fill r28=[r2],16
> + ld8.fill r29=[r3],16
> + ;;
> + ld8.fill r30=[r2],16
> + ld8.fill r31=[r3],16
> + ;;
> + mov ar.unat=r15
> + ;;
> + bsw.0
> + ;;
> + br.ret.sptk.many b0
> +END(vmx_asm_bsw0)
> +
> +//r13 ->vcpu
> +//call with psr.bn = 0
> +GLOBAL_ENTRY(vmx_asm_bsw1)
> + mov r15=ar.unat
> + ;;
> + adds r14=IA64_VPD_BASE_OFFSET,r13
> + ;;
> + ld8 r14=[r14]
> + bsw.1
> + ;;
> + adds r2= IA64_VPD_VB0REG_OFFSET, r14
> + adds r3= IA64_VPD_VB0REG_OFFSET+8, r14
> + ;;
> + .mem.offset 0,0; st8.spill [r2]=r16,16
> + .mem.offset 8,0; st8.spill [r3]=r17,16
> + ;;
> + .mem.offset 0,0; st8.spill [r2]=r18,16
> + .mem.offset 8,0; st8.spill [r3]=r19,16
> + ;;
> + .mem.offset 0,0; st8.spill [r2]=r20,16
> + .mem.offset 8,0; st8.spill [r3]=r21,16
> + ;;
> + .mem.offset 0,0; st8.spill [r2]=r22,16
> + .mem.offset 8,0; st8.spill [r3]=r23,16
> + ;;
> + .mem.offset 0,0; st8.spill [r2]=r24,16
> + .mem.offset 8,0; st8.spill [r3]=r25,16
> + ;;
> + .mem.offset 0,0; st8.spill [r2]=r26,16
> + .mem.offset 8,0; st8.spill [r3]=r27,16
> + ;;
> + .mem.offset 0,0; st8.spill [r2]=r28,16
> + .mem.offset 8,0; st8.spill [r3]=r29,16
> + ;;
> + .mem.offset 0,0; st8.spill [r2]=r30,16
> + .mem.offset 8,0; st8.spill [r3]=r31,16
> + ;;
> + mov r9=ar.unat
> + adds r8=IA64_VPD_VB0NAT_OFFSET, r14
> + ;;
> + st8 [r8]=r9
> + adds r8=IA64_VPD_VB1NAT_OFFSET, r14
> + ;;
> + ld8 r9=[r8]
> + adds r2= IA64_VPD_VB1REG_OFFSET, r14
> + adds r3= IA64_VPD_VB1REG_OFFSET+8, r14
> + ;;
> + mov ar.unat=r9
> + ;;
> + ld8.fill r16=[r2],16
> + ld8.fill r17=[r3],16
> + ;;
> + ld8.fill r18=[r2],16
> + ld8.fill r19=[r3],16
> + ;;
> + ld8.fill r20=[r2],16
> + ld8.fill r21=[r3],16
> + ;;
> + ld8.fill r22=[r2],16
> + ld8.fill r23=[r3],16
> + ;;
> + ld8.fill r24=[r2],16
> + ld8.fill r25=[r3],16
> + ;;
> + ld8.fill r26=[r2],16
> + ld8.fill r27=[r3],16
> + ;;
> + ld8.fill r28=[r2],16
> + ld8.fill r29=[r3],16
> + ;;
> + ld8.fill r30=[r2],16
> + ld8.fill r31=[r3],16
> + ;;
> + mov ar.unat=r15
> + ;;
> + bsw.0
> + ;;
> + br.ret.sptk.many b0
> +END(vmx_asm_bsw1)
> +
> +
> +// rfi
> +ENTRY(vmx_asm_rfi)
> + adds r18=IA64_VPD_BASE_OFFSET,r21
> + ;;
> + ld8 r18=[r18]
> + ;;
> + adds r26=IA64_VPD_VIFS_OFFSET,r18
> + ;;
> + ld8 r26=[r26]
> + ;;
> + tbit.z p6,p0=r26,63
> + (p6) br.cond.dptk.few vmx_asm_rfi_1
> + ;;
> + //if vifs.v=1 desert current register frame
> + alloc r27=ar.pfs,0,0,0,0
> + ;;
> +vmx_asm_rfi_1:
> + adds r26=IA64_VPD_VHPI_OFFSET,r18
> + ;;
> + ld8 r26=[r26]
> + ;;
> + cmp.ne p6,p0=r26,r0
> + (p6) br.cond.dpnt.many vmx_virtualization_fault_back
> + ;;
> + VMX_VIRT_SAVE
> + ;;
> + mov out0=r21
> + movl r14=ia64_leave_hypervisor_virt
> + ;;
> + mov rp=r14
> + br.call.sptk.many b6=vmx_vcpu_rfi_fast
> +END(vmx_asm_rfi)
> +
>
> //mov r1=ar3 (only itc is virtualized)
> GLOBAL_ENTRY(vmx_asm_mov_from_ar)
> -#ifndef ACCE_MOV_FROM_AR
> - br.many vmx_virtualization_fault_back
> -#endif
Now ACCE_MOV_FROM_AR doesn't seem to make sense.
Please remove the corresponding #define.
> add r18=VCPU_VTM_OFFSET_OFS,r21
> add r16=VCPU_VTM_LAST_ITC_OFS,r21
> extr.u r17=r25,6,7
> @@ -128,9 +657,6 @@ END(vmx_asm_mov_from_ar)
>
> // mov r1=rr[r3]
> GLOBAL_ENTRY(vmx_asm_mov_from_rr)
> -#ifndef ACCE_MOV_FROM_RR
> - br.many vmx_virtualization_fault_back
> -#endif
ditto.
> extr.u r16=r25,20,7
> extr.u r17=r25,6,7
> movl r20=asm_mov_from_reg
> @@ -162,270 +688,96 @@ GLOBAL_ENTRY(vmx_asm_mov_to_rr)
> #ifndef ACCE_MOV_TO_RR
> br.many vmx_virtualization_fault_back
> #endif
> - add r22=IA64_VCPU_RID_BITS_OFFSET,r21
> extr.u r16=r25,20,7 // r3
> extr.u r17=r25,13,7 // r2
> - ;;
> movl r20=asm_mov_from_reg
> ;;
> adds r30=vmx_asm_mov_to_rr_back_1-asm_mov_from_reg,r20
> shladd r16=r16,4,r20 // get r3
> - mov r18=b0 // save b0
> ;;
> - add r27=VCPU_VRR0_OFS,r21
> mov b0=r16
> br.many b0
> ;;
> vmx_asm_mov_to_rr_back_1:
> adds r30=vmx_asm_mov_to_rr_back_2-asm_mov_from_reg,r20
> - shr.u r23=r19,61 // get RR #
> + shr.u r16=r19,61 // get RR #
> + ;;
> + //if rr7, go back
> + cmp.eq p6,p0=7,r16
> + mov b0=r23 // restore b0
> + (p6) br.cond.dpnt.many vmx_virtualization_fault_back
> + ;;
> + mov r16=r19
> shladd r17=r17,4,r20 // get r2
> ;;
> - //if rr7, go back
> - cmp.eq p6,p0=7,r23
> - mov b0=r18 // restore b0
> - (p6) br.cond.dpnt.many vmx_virtualization_fault_back
> - ;;
> - mov r28=r19 // save r3
> mov b0=r17
> br.many b0
> vmx_asm_mov_to_rr_back_2:
> - adds r30=vmx_resume_to_guest-asm_mov_from_reg,r20
> - shladd r27=r23,3,r27 // address of VRR
> + mov r17=r19 // get value
> ;;
> - ld1 r22=[r22] // Load rid_bits from domain
> - mov b0=r18 // restore b0
> - adds r16=IA64_VCPU_STARTING_RID_OFFSET,r21
> + VMX_VIRT_SAVE
> ;;
> - ld4 r16=[r16] // load starting_rid
> - extr.u r17=r19,8,24 // Extract RID
> + mov out0=r21
> + mov out1=r16
> + mov out2=r17
> + movl r14=ia64_leave_hypervisor_virt
> ;;
> - shr r17=r17,r22 // Shift out used bits
> - shl r16=r16,8
> - ;;
> - add r20=r19,r16
> - cmp.ne p6,p0=0,r17 // If reserved RID bits are set, use C fall
> back.
> - (p6) br.cond.dpnt.many vmx_virtualization_fault_back
> - ;; //mangling rid 1 and 3
> - extr.u r16=r20,8,8
> - extr.u r17=r20,24,8
> - mov r24=r18 // saved b0 for resume
> - ;;
> - extr.u r18=r20,2,6 // page size
> - dep r20=r16,r20,24,8
> - mov b0=r30
> - ;;
> - dep r20=r17,r20,8,8
> - ;; //set ve 1
> - dep r20=-1,r20,0,1
> - // If ps > PAGE_SHIFT, use PAGE_SHIFT
> - cmp.lt p6,p0=PAGE_SHIFT,r18
> - ;;
> - (p6) mov r18=PAGE_SHIFT
> - ;;
> - (p6) dep r20=r18,r20,2,6
> - ;;
> - st8 [r27]=r19 // Write to vrr.
> - // Write to save_rr if rr=0 or rr=4.
> - cmp.eq p6,p0=0,r23
> - ;;
> - cmp.eq.or p6,p0=4,r23
> - ;;
> - adds r16=IA64_VCPU_MMU_MODE_OFFSET,r21
> - (p6) adds r17=IA64_VCPU_META_SAVED_RR0_OFFSET,r21
> - ;;
> - ld1 r16=[r16]
> - cmp.eq p7,p0=r0,r0
> - (p6) shladd r17=r23,1,r17
> - ;;
> - (p6) st8 [r17]=r20
> - (p6) cmp.eq p7,p0=VMX_MMU_VIRTUAL,r16 // Set physical rr if in virt mode
> - ;;
> - (p7) mov rr[r28]=r20
> - br.many b0
> + mov rp=r14
> + br.call.sptk.many b6=vmx_vcpu_set_rr
> END(vmx_asm_mov_to_rr)
>
>
> -//rsm
> +//rsm 25
> GLOBAL_ENTRY(vmx_asm_rsm)
> -#ifndef ACCE_RSM
> - br.many vmx_virtualization_fault_back
> -#endif
ditto.
> - mov r23=r31
> - add r16=IA64_VPD_BASE_OFFSET,r21
> extr.u r26=r25,6,21 // Imm21
> extr.u r27=r25,31,2 // I2d
> ;;
> - ld8 r16=[r16]
> extr.u r28=r25,36,1 // I
> dep r26=r27,r26,21,2
> ;;
> - add r17=VPD_VPSR_START_OFFSET,r16
> //r18 is imm24
> - dep r18=r28,r26,23,1
> + dep r16=r28,r26,23,1
> ;;
> - //sync read
> - mov r25=r16
> - movl r24=vmx_asm_rsm_sync_read_return
> - mov r20=b0
> - br.sptk.many vmx_vps_sync_read
> + VMX_VIRT_SAVE
> ;;
> -vmx_asm_rsm_sync_read_return:
> - ld8 r26=[r17]
> - // xenoprof
> - // Don't change mPSR.pp.
> - // It is manipulated by xenoprof.
> - movl r28=IA64_PSR_IC+IA64_PSR_I+IA64_PSR_DT+IA64_PSR_SI+IA64_PSR_PP
> + mov out0=r21
> + mov out1=r16
> + movl r14=ia64_leave_hypervisor_virt
> + ;;
> + mov rp=r14
> + br.call.sptk.many b6=vmx_vcpu_rsm_fast
> +END(vmx_asm_rsm)
>
> - sub r27=-1,r18 // ~imm24
> +
> +//ssm 24
> +GLOBAL_ENTRY(vmx_asm_ssm)
> + adds r18=IA64_VPD_BASE_OFFSET,r21
> ;;
> - or r28=r27,r28 // Keep IC,I,DT,SI
> - and r19=r26,r27 // Update vpsr
> - ;;
> - st8 [r17]=r19
> - mov r24=cr.ipsr
> - ;;
> - and r24=r24,r28 // Update ipsr
> - adds r27=IA64_VCPU_FP_PSR_OFFSET,r21
> - ;;
> - ld8 r27=[r27]
> - ;;
> - tbit.nz p8,p0=r27,IA64_PSR_DFH_BIT
> - ;;
> - (p8) dep r24=-1,r24,IA64_PSR_DFH_BIT,1 // Keep dfh
> - ;;
> - mov cr.ipsr=r24
> - //sync write
> - mov r25=r16
> - movl r24=vmx_asm_rsm_sync_write_return
> - br.sptk.many vmx_vps_sync_write
> - ;;
> -vmx_asm_rsm_sync_write_return:
> - add r29=IA64_VCPU_MMU_MODE_OFFSET,r21
> - ;;
> - ld1 r27=[r29]
> - ;;
> - cmp.ne p6,p0=VMX_MMU_VIRTUAL,r27
> - ;;
> - tbit.z.or p6,p0=r18,IA64_PSR_DT_BIT
> - (p6) br.dptk vmx_asm_rsm_out
> - // DT not cleared or already in phy mode
> - ;;
> - // Switch to meta physical mode D.
> - add r26=IA64_VCPU_META_RID_D_OFFSET,r21
> - mov r27=VMX_MMU_PHY_D
> + ld8 r18=[r18]
> + ;;
> + adds r26=IA64_VPD_VHPI_OFFSET,r18
> ;;
> ld8 r26=[r26]
> - st1 [r29]=r27
> - dep.z r28=4,61,3
> ;;
> - mov rr[r0]=r26
> + cmp.ne p6,p0=r26,r0
> + (p6) br.cond.dpnt.many vmx_virtualization_fault_back
> ;;
> - mov rr[r28]=r26
> - ;;
> - srlz.d
> -vmx_asm_rsm_out:
> - mov r31=r23
> - mov r24=r20
> - br.many vmx_resume_to_guest
> -END(vmx_asm_rsm)
> -
> -
> -//ssm
> -GLOBAL_ENTRY(vmx_asm_ssm)
> -#ifndef ACCE_SSM
> - br.many vmx_virtualization_fault_back
> -#endif
> - mov r23=r31
> - add r16=IA64_VPD_BASE_OFFSET,r21
> extr.u r26=r25,6,21
> extr.u r27=r25,31,2
> ;;
> - ld8 r16=[r16]
> extr.u r28=r25,36,1
> dep r26=r27,r26,21,2
> ;; //r18 is imm24
> - dep r18=r28,r26,23,1
> - ;;
> - //sync read
> - mov r25=r16
> - movl r24=vmx_asm_ssm_sync_read_return
> - mov r20=b0
> - br.sptk.many vmx_vps_sync_read
> + dep r16=r28,r26,23,1
> + ;; ;;
> + VMX_VIRT_SAVE
> ;;
> -vmx_asm_ssm_sync_read_return:
> - add r27=VPD_VPSR_START_OFFSET,r16
> + mov out0=r21
> + mov out1=r16
> + movl r14=ia64_leave_hypervisor_virt
> ;;
> - ld8 r17=[r27] //r17 old vpsr
> - dep r28=0,r18,IA64_PSR_PP_BIT,1 // For xenoprof
> - // Don't change mPSR.pp
> - // It is maintained by xenoprof.
> - ;;
> - or r19=r17,r18 //r19 new vpsr
> - ;;
> - st8 [r27]=r19 // update vpsr
> - mov r24=cr.ipsr
> - ;;
> - or r24=r24,r28
> - ;;
> - mov cr.ipsr=r24
> - //sync_write
> - mov r25=r16
> - movl r24=vmx_asm_ssm_sync_write_return
> - br.sptk.many vmx_vps_sync_write
> - ;;
> -vmx_asm_ssm_sync_write_return:
> - add r29=IA64_VCPU_MMU_MODE_OFFSET,r21
> - movl r28=IA64_PSR_DT+IA64_PSR_RT+IA64_PSR_IT
> - ;;
> - ld1 r30=[r29] // mmu_mode
> - ;;
> - and r27=r28,r19
> - cmp.eq p6,p0=VMX_MMU_VIRTUAL,r30
> - ;;
> - cmp.ne.or p6,p0=r28,r27 // (vpsr & (it+dt+rt)) /= (it+dt+rt) ie stay in
> phy
> - (p6) br.dptk vmx_asm_ssm_1
> - ;;
> - add r26=IA64_VCPU_META_SAVED_RR0_OFFSET,r21
> - add r27=IA64_VCPU_META_SAVED_RR0_OFFSET+8,r21
> - mov r30=VMX_MMU_VIRTUAL
> - ;;
> - ld8 r26=[r26]
> - ld8 r27=[r27]
> - st1 [r29]=r30
> - dep.z r28=4,61,3
> - ;;
> - mov rr[r0]=r26
> - ;;
> - mov rr[r28]=r27
> - ;;
> - srlz.d
> - ;;
> -vmx_asm_ssm_1:
> - tbit.nz p6,p0=r17,IA64_PSR_I_BIT
> - ;;
> - tbit.z.or p6,p0=r19,IA64_PSR_I_BIT
> - (p6) br.dptk vmx_asm_ssm_out
> - ;;
> - add r29=VPD_VTPR_START_OFFSET,r16
> - add r30=VPD_VHPI_START_OFFSET,r16
> - ;;
> - ld8 r29=[r29]
> - ld8 r30=[r30]
> - ;;
> - extr.u r17=r29,4,4
> - extr.u r18=r29,16,1
> - ;;
> - dep r17=r18,r17,4,1
> - mov r31=r23
> - mov b0=r20
> - ;;
> - cmp.gt p6,p0=r30,r17
> - (p6) br.dpnt.few vmx_asm_dispatch_vexirq
> -vmx_asm_ssm_out:
> - mov r31=r23
> - mov r24=r20
> - br.many vmx_resume_to_guest
> + mov rp=r14
> + br.call.sptk.many b6=vmx_vcpu_ssm_fast
> END(vmx_asm_ssm)
>
>
> @@ -434,141 +786,41 @@ GLOBAL_ENTRY(vmx_asm_mov_to_psr)
> #ifndef ACCE_MOV_TO_PSR
> br.many vmx_virtualization_fault_back
> #endif
> - mov r23=r31
> - add r16=IA64_VPD_BASE_OFFSET,r21
> extr.u r26=r25,13,7 //r2
> + movl r27=asm_mov_from_reg
> ;;
> - ld8 r16=[r16]
> - movl r24=asm_mov_from_reg
> - ;;
> - adds r30=vmx_asm_mov_to_psr_back-asm_mov_from_reg,r24
> - shladd r26=r26,4,r24
> - mov r20=b0
> + adds r30=vmx_asm_mov_to_psr_back-asm_mov_from_reg,r27
> + shladd r26=r26,4,r27
> ;;
> mov b0=r26
> br.many b0
> ;;
> vmx_asm_mov_to_psr_back:
> - //sync read
> - mov r25=r16
> - movl r24=vmx_asm_mov_to_psr_sync_read_return
> - br.sptk.many vmx_vps_sync_read
> + adds r18=IA64_VPD_BASE_OFFSET,r21
> + tbit.nz p6,p0 = r19, IA64_PSR_I_BIT
> ;;
> -vmx_asm_mov_to_psr_sync_read_return:
> - add r27=VPD_VPSR_START_OFFSET,r16
> - ;;
> - ld8 r17=[r27] // r17 old vpsr
> - dep r19=0,r19,32,32 // Clear bits 32-63
> - ;;
> - dep r18=0,r17,0,32
> - ;;
> - or r18=r18,r19 //r18 new vpsr
> - ;;
> - st8 [r27]=r18 // set vpsr
> - //sync write
> - mov r25=r16
> - movl r24=vmx_asm_mov_to_psr_sync_write_return
> - br.sptk.many vmx_vps_sync_write
> - ;;
> -vmx_asm_mov_to_psr_sync_write_return:
> - add r22=IA64_VCPU_MMU_MODE_OFFSET,r21
> - movl r28=IA64_PSR_DT+IA64_PSR_RT+IA64_PSR_IT
> - ;;
> - and r27=r28,r18
> - and r29=r28,r17
> - ;;
> - cmp.eq p5,p0=r29,r27 // (old_vpsr & (dt+rt+it)) == (new_vpsr &
> (dt+rt+it))
> - cmp.eq p6,p7=r28,r27 // (new_vpsr & (dt+rt+it)) == (dt+rt+it)
> - (p5) br.many vmx_asm_mov_to_psr_1 // no change
> - ;;
> - //virtual to physical D
> - (p7) add r26=IA64_VCPU_META_RID_D_OFFSET,r21
> - (p7) add r27=IA64_VCPU_META_RID_D_OFFSET,r21
> - (p7) mov r30=VMX_MMU_PHY_D
> - ;;
> - //physical to virtual
> - (p6) add r26=IA64_VCPU_META_SAVED_RR0_OFFSET,r21
> - (p6) add r27=IA64_VCPU_META_SAVED_RR0_OFFSET+8,r21
> - (p6) mov r30=VMX_MMU_VIRTUAL
> + ld8 r18=[r18]
> + ;;
> + adds r26=IA64_VPD_VHPI_OFFSET,r18
> ;;
> ld8 r26=[r26]
> - ld8 r27=[r27]
> - st1 [r22]=r30
> - dep.z r28=4,61,3
> ;;
> - mov rr[r0]=r26
> + // if enable interrupt and vhpi has value, return
> + cmp.ne.and p6,p0=r26,r0
> + (p6) br.cond.dpnt.many vmx_virtualization_fault_back
> ;;
> - mov rr[r28]=r27
> + mov r16=r19
> ;;
> - srlz.d
> + VMX_VIRT_SAVE
> ;;
> -vmx_asm_mov_to_psr_1:
> - mov r24=cr.ipsr
> - movl r28=IA64_PSR_IC+IA64_PSR_I+IA64_PSR_DT+IA64_PSR_SI+IA64_PSR_RT
> + mov out0=r21
> + mov out1=r16
> + movl r14=ia64_leave_hypervisor_virt
> ;;
> - tbit.nz p7,p0=r24,IA64_PSR_PP_BIT // For xenoprof
> - or r27=r19,r28
> - dep r24=0,r24,0,32
> - ;;
> - add r24=r27,r24
> - ;;
> - adds r27=IA64_VCPU_FP_PSR_OFFSET,r21
> - (p7) dep r24=-1,r24,IA64_PSR_PP_BIT,1 // For xenoprof
> - // Dom't change mPSR.pp
> - // It is maintaned by
> xenoprof
> - ;;
> - ld8 r27=[r27]
> - ;;
> - tbit.nz p8,p0=r27,IA64_PSR_DFH_BIT
> - ;;
> - (p8) dep r24=-1,r24,IA64_PSR_DFH_BIT,1
> - ;;
> - mov cr.ipsr=r24
> - tbit.nz p6,p0=r17,IA64_PSR_I_BIT
> - ;;
> - tbit.z.or p6,p0=r18,IA64_PSR_I_BIT
> - (p6) br.dpnt.few vmx_asm_mov_to_psr_out
> - ;;
> - add r29=VPD_VTPR_START_OFFSET,r16
> - add r30=VPD_VHPI_START_OFFSET,r16
> - ;;
> - ld8 r29=[r29]
> - ld8 r30=[r30]
> - ;;
> - extr.u r17=r29,4,4
> - extr.u r18=r29,16,1
> - ;;
> - dep r17=r18,r17,4,1
> - mov r31=r23
> - mov b0=r20
> - ;;
> - cmp.gt p6,p0=r30,r17
> - (p6) br.dpnt.few vmx_asm_dispatch_vexirq
> -vmx_asm_mov_to_psr_out:
> - mov r31=r23
> - mov r24=r20
> - br.many vmx_resume_to_guest
> + mov rp=r14
> + br.call.sptk.many b6=vmx_vcpu_mov_to_psr_fast
> END(vmx_asm_mov_to_psr)
>
> -
> -ENTRY(vmx_asm_dispatch_vexirq)
> -//increment iip
> - mov r16=cr.ipsr
> - ;;
> - extr.u r17=r16,IA64_PSR_RI_BIT,2
> - tbit.nz p6,p7=r16,IA64_PSR_RI_BIT+1
> - ;;
> - (p6) mov r18=cr.iip
> - (p6) mov r17=r0
> - (p7) add r17=1,r17
> - ;;
> - (p6) add r18=0x10,r18
> - dep r16=r17,r16,IA64_PSR_RI_BIT,2
> - ;;
> - (p6) mov cr.iip=r18
> - mov cr.ipsr=r16
> - br.many vmx_dispatch_vexirq
> -END(vmx_asm_dispatch_vexirq)
>
> // thash r1=r3
> // TODO: add support when pta.vf = 1
> @@ -633,6 +885,8 @@ vmx_asm_thash_back1:
> mov b0=r17
> br.many b0
> END(vmx_asm_thash)
> +
> +
>
> #define MOV_TO_REG0 \
> {; \
> @@ -887,20 +1141,21 @@ END(asm_mov_from_reg)
> * parameter:
> * r31: pr
> * r24: b0
> + * p2: whether increase IP
> + * p3: whether check vpsr.ic
> */
> ENTRY(vmx_resume_to_guest)
> - mov r16=cr.ipsr
> - ;;
> + // ip ++
> + (p2) mov r16=cr.ipsr
> + (p2)dep.z r30=1,IA64_PSR_RI_BIT,1
> adds r19=IA64_VPD_BASE_OFFSET,r21
> - extr.u r17=r16,IA64_PSR_RI_BIT,2
> ;;
> ld8 r25=[r19]
> - add r17=1,r17
> + (p2) add r16=r30,r16
> ;;
> + (p2) mov cr.ipsr=r16
> adds r19= VPD_VPSR_START_OFFSET,r25
> - dep r16=r17,r16,IA64_PSR_RI_BIT,2
> ;;
> - mov cr.ipsr=r16
> ld8 r19=[r19]
> ;;
> mov r23=r31
> diff -r f2457c7aff8d xen/arch/ia64/vmx/vmx_ivt.S
> --- a/xen/arch/ia64/vmx/vmx_ivt.S Fri Apr 25 20:13:52 2008 +0900
> +++ b/xen/arch/ia64/vmx/vmx_ivt.S Thu May 08 16:23:42 2008 +0800
> @@ -967,21 +967,13 @@ ENTRY(vmx_virtualization_fault)
> ENTRY(vmx_virtualization_fault)
> // VMX_DBG_FAULT(37)
> mov r31=pr
> + movl r30 = virtualization_fault_table
> + mov r23=b0
> ;;
> - cmp.eq p6,p0=EVENT_MOV_FROM_AR,r24
> - cmp.eq p7,p0=EVENT_MOV_FROM_RR,r24
> - cmp.eq p8,p0=EVENT_MOV_TO_RR,r24
> - cmp.eq p9,p0=EVENT_RSM,r24
> - cmp.eq p10,p0=EVENT_SSM,r24
> - cmp.eq p11,p0=EVENT_MOV_TO_PSR,r24
> - cmp.eq p12,p0=EVENT_THASH,r24
> - (p6) br.dptk.many vmx_asm_mov_from_ar
> - (p7) br.dptk.many vmx_asm_mov_from_rr
> - (p8) br.dptk.many vmx_asm_mov_to_rr
> - (p9) br.dptk.many vmx_asm_rsm
> - (p10) br.dptk.many vmx_asm_ssm
> - (p11) br.dptk.many vmx_asm_mov_to_psr
> - (p12) br.dptk.many vmx_asm_thash
> + shladd r30=r24,4,r30
> + ;;
> + mov b0=r30
> + br.sptk.many b0
> ;;
> vmx_virtualization_fault_back:
> mov r19=37
> @@ -990,23 +982,6 @@ vmx_virtualization_fault_back:
> ;;
> st8 [r16] = r24
> st8 [r17] = r25
> - ;;
> - cmp.ne p6,p0=EVENT_RFI, r24
> - (p6) br.sptk vmx_dispatch_virtualization_fault
> - ;;
> - adds r18=IA64_VPD_BASE_OFFSET,r21
> - ;;
> - ld8 r18=[r18]
> - ;;
> - adds r18=IA64_VPD_VIFS_OFFSET,r18
> - ;;
> - ld8 r18=[r18]
> - ;;
> - tbit.z p6,p0=r18,63
> - (p6) br.sptk vmx_dispatch_virtualization_fault
> - ;;
> - //if vifs.v=1 desert current register frame
> - alloc r18=ar.pfs,0,0,0,0
> br.sptk vmx_dispatch_virtualization_fault
> END(vmx_virtualization_fault)
>
> diff -r f2457c7aff8d xen/arch/ia64/vmx/vmx_phy_mode.c
> --- a/xen/arch/ia64/vmx/vmx_phy_mode.c Fri Apr 25 20:13:52 2008 +0900
> +++ b/xen/arch/ia64/vmx/vmx_phy_mode.c Thu May 08 16:23:42 2008 +0800
> @@ -252,8 +252,8 @@ switch_mm_mode(VCPU *vcpu, IA64_PSR old_
> switch_to_virtual_rid(vcpu);
> break;
> case SW_SELF:
> - printk("Switch to self-0x%lx!!! MM mode doesn't change...\n",
> - old_psr.val);
> +// printk("Switch to self-0x%lx!!! MM mode doesn't change...\n",
> +// old_psr.val);
> break;
> case SW_NOP:
> // printk("No action required for mode transition: (0x%lx ->
> 0x%lx)\n",
What's the purpose here.
Anyway if you want this part, please create another patch.
> diff -r f2457c7aff8d xen/arch/ia64/vmx/vmx_vcpu.c
> --- a/xen/arch/ia64/vmx/vmx_vcpu.c Fri Apr 25 20:13:52 2008 +0900
> +++ b/xen/arch/ia64/vmx/vmx_vcpu.c Thu May 08 16:23:42 2008 +0800
> @@ -172,11 +172,6 @@ IA64FAULT vmx_vcpu_set_rr(VCPU *vcpu, u6
> {
> u64 rrval;
>
> - if (unlikely(is_reserved_rr_rid(vcpu, val))) {
> - gdprintk(XENLOG_DEBUG, "use of invalid rrval %lx\n", val);
> - return IA64_RSVDREG_FAULT;
> - }
> -
> VMX(vcpu,vrr[reg>>VRN_SHIFT]) = val;
> switch((u64)(reg>>VRN_SHIFT)) {
> case VRN7:
ditto.
--
yamahata
_______________________________________________
Xen-ia64-devel mailing list
Xen-ia64-devel@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-ia64-devel
|
Home