[PATCH 08/27] Add SLB switching code for entry/exit

[Alexander Graf]

Am 02.11.2009 um 00:23 schrieb Michael Neuling <mikey at neuling.org>:

>> This is the really low level of guest entry/exit code.
>>
>> Book3s_64 has an SLB, which stores all ESID -> VSID mappings we're
>> currently aware of.
>>
>> The segments in the guest differ from the ones on the host, so we  
>> need
>> to switch the SLB to tell the MMU that we're in a new context.
>>
>> So we store a shadow of the guest's SLB in the PACA, switch to that  
>> on
>> entry and only restore bolted entries on exit, leaving the rest to  
>> the
>> Linux SLB fault handler.
>>
>> That way we get a really clean way of switching the SLB.
>>
>> Signed-off-by: Alexander Graf <agraf at suse.de>
>> ---
>> arch/powerpc/kvm/book3s_64_slb.S |  277 ++++++++++++++++++++++++++++ 
>> ++++++++
> ++
>> 1 files changed, 277 insertions(+), 0 deletions(-)
>> create mode 100644 arch/powerpc/kvm/book3s_64_slb.S
>>
>> diff --git a/arch/powerpc/kvm/book3s_64_slb.S b/arch/powerpc/kvm/ 
>> book3s_64_sl
> b.S
>> new file mode 100644
>> index 0000000..00a8367
>> --- /dev/null
>> +++ b/arch/powerpc/kvm/book3s_64_slb.S
>> @@ -0,0 +1,277 @@
>> +/*
>> + * This program is free software; you can redistribute it and/or  
>> modify
>> + * it under the terms of the GNU General Public License, version  
>> 2, as
>> + * published by the Free Software Foundation.
>> + *
>> + * This program is distributed in the hope that it will be useful,
>> + * but WITHOUT ANY WARRANTY; without even the implied warranty of
>> + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
>> + * GNU General Public License for more details.
>> + *
>> + * You should have received a copy of the GNU General Public License
>> + * along with this program; if not, write to the Free Software
>> + * Foundation, 51 Franklin Street, Fifth Floor, Boston, MA   
>> 02110-1301, USA.
>> + *
>> + * Copyright SUSE Linux Products GmbH 2009
>> + *
>> + * Authors: Alexander Graf <agraf at suse.de>
>> + */
>> +
>> +/ 
>> *** 
>> *** 
>> *********************************************************************
> ***
>> + *
>  *
>> + *                               Entry code
>  *
>> + *
>  *
>> +  
>> *** 
>> *** 
>> *********************************************************************
> **/
>> +
>> +.global kvmppc_handler_trampoline_enter
>> +kvmppc_handler_trampoline_enter:
>> +
>> +    /* Required state:
>> +     *
>> +     * MSR = ~IR|DR
>> +     * R13 = PACA
>> +     * R9 = guest IP
>> +     * R10 = guest MSR
>> +     * R11 = free
>> +     * R12 = free
>> +     * PACA[PACA_EXMC + EX_R9] = guest R9
>> +     * PACA[PACA_EXMC + EX_R10] = guest R10
>> +     * PACA[PACA_EXMC + EX_R11] = guest R11
>> +     * PACA[PACA_EXMC + EX_R12] = guest R12
>> +     * PACA[PACA_EXMC + EX_R13] = guest R13
>> +     * PACA[PACA_EXMC + EX_CCR] = guest CR
>> +     * PACA[PACA_EXMC + EX_R3] = guest XER
>> +     */
>> +
>> +    mtsrr0    r9
>> +    mtsrr1    r10
>> +
>> +    mtspr    SPRN_SPRG_SCRATCH0, r0
>> +
>> +    /* Remove LPAR shadow entries */
>> +
>> +#if SLB_NUM_BOLTED == 3
>
> You could alternatively check the persistent entry in the slb_shawdow
> buffer.  This would give you a run time check.  Not sure what's best
> though.

Well we're in the hot path here, so anything using as few registers as  
possible and being simple is the best :-). I'd guess the more we are  
clever at compile time the better.

>
>
>> +
>> +    ld    r12, PACA_SLBSHADOWPTR(r13)
>> +    ld    r10, 0x10(r12)
>> +    ld    r11, 0x18(r12)
>
> Can you define something in asm-offsets.c for these magic constants  
> 0x10
> and 0x18.  Similarly below.
>
>> +    /* Invalid? Skip. */
>> +    rldicl. r0, r10, 37, 63
>> +    beq    slb_entry_skip_1
>> +    xoris    r9, r10, SLB_ESID_V at h
>> +    std    r9, 0x10(r12)
>> +slb_entry_skip_1:
>> +    ld    r9, 0x20(r12)
>> +    /* Invalid? Skip. */
>> +    rldicl. r0, r9, 37, 63
>> +    beq    slb_entry_skip_2
>> +    xoris    r9, r9, SLB_ESID_V at h
>> +    std    r9, 0x20(r12)
>> +slb_entry_skip_2:
>> +    ld    r9, 0x30(r12)
>> +    /* Invalid? Skip. */
>> +    rldicl. r0, r9, 37, 63
>> +    beq    slb_entry_skip_3
>> +    xoris    r9, r9, SLB_ESID_V at h
>> +    std    r9, 0x30(r12)
>
> Can these 3 be made into a macro?

Phew - dynamically generating jump points sounds rather hard. I can  
give it a try...

>
>> +slb_entry_skip_3:
>> +
>> +#else
>> +#error unknown number of bolted entries
>> +#endif
>> +
>> +    /* Flush SLB */
>> +
>> +    slbia
>> +
>> +    /* r0 = esid & ESID_MASK */
>> +    rldicr  r10, r10, 0, 35
>> +    /* r0 |= CLASS_BIT(VSID) */
>> +    rldic   r12, r11, 56 - 36, 36
>> +    or      r10, r10, r12
>> +    slbie    r10
>> +
>> +    isync
>> +
>> +    /* Fill SLB with our shadow */
>> +
>> +    lbz    r12, PACA_KVM_SLB_MAX(r13)
>> +    mulli    r12, r12, 16
>> +    addi    r12, r12, PACA_KVM_SLB
>> +    add    r12, r12, r13
>> +
>> +    /* for (r11 = kvm_slb; r11 < kvm_slb + kvm_slb_size;  
>> r11+=slb_entry) */
>> +    li    r11, PACA_KVM_SLB
>> +    add    r11, r11, r13
>> +
>> +slb_loop_enter:
>> +
>> +    ld    r10, 0(r11)
>> +
>> +    rldicl. r0, r10, 37, 63
>> +    beq    slb_loop_enter_skip
>> +
>> +    ld    r9, 8(r11)
>> +    slbmte    r9, r10
>
> If you're updating the first 3 slbs, you need to make sure the slb
> shadow is updated at the same time

Well - what happens if we don't? We'd get a segment fault when phyp  
stole our entry! So what? Let it fault, see the mapping is already  
there and get back in again :-).

> (BTW dumb question: can we run this
> under PHYP?)

Yes, I tested it on bare metal, phyp and a PS3.


Alex