From: Michael Neuling <mikey@neuling.org>
To: Alexander Graf <agraf@suse.de>
Cc: kvm@vger.kernel.org, Kevin Wolf <kwolf@redhat.com>,
Arnd Bergmann <arnd@arndb.de>,
Hollis Blanchard <hollisb@us.ibm.com>,
Marcelo Tosatti <mtosatti@redhat.com>,
kvm-ppc <kvm-ppc@vger.kernel.org>,
linuxppc-dev@ozlabs.org, Avi Kivity <avi@redhat.com>,
bphilips@suse.de, Olof Johansson <olof@lixom.net>
Subject: Re: [PATCH 08/27] Add SLB switching code for entry/exit
Date: Mon, 02 Nov 2009 10:23:47 +1100 [thread overview]
Message-ID: <6695.1257117827@neuling.org> (raw)
In-Reply-To: <1256917647-6200-9-git-send-email-agraf@suse.de>
> 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@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@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.
> +
> + 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@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@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@h
> + std r9, 0x30(r12)
Can these 3 be made into a macro?
> +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 (BTW dumb question: can we run this
under PHYP?)
> +
> +slb_loop_enter_skip:
> + addi r11, r11, 16
> + cmpd cr0, r11, r12
> + blt slb_loop_enter
> +
> +slb_do_enter:
> +
> + /* Enter guest */
> +
> + mfspr r0, SPRN_SPRG_SCRATCH0
> +
> + ld r9, (PACA_EXMC+EX_R9)(r13)
> + ld r10, (PACA_EXMC+EX_R10)(r13)
> + ld r12, (PACA_EXMC+EX_R12)(r13)
> +
> + lwz r11, (PACA_EXMC+EX_CCR)(r13)
> + mtcr r11
> +
> + ld r11, (PACA_EXMC+EX_R3)(r13)
> + mtxer r11
> +
> + ld r11, (PACA_EXMC+EX_R11)(r13)
> + ld r13, (PACA_EXMC+EX_R13)(r13)
> +
> + RFI
> +kvmppc_handler_trampoline_enter_end:
> +
> +
> +
> +/***************************************************************************
***
> + *
*
> + * Exit code
*
> + *
*
> + ***************************************************************************
**/
> +
> +.global kvmppc_handler_trampoline_exit
> +kvmppc_handler_trampoline_exit:
> +
> + /* Register usage at this point:
> + *
> + * SPRG_SCRATCH0 = guest R13
> + * R01 = host R1
> + * R02 = host R2
> + * R10 = guest PC
> + * R11 = guest MSR
> + * R12 = exit handler id
> + * R13 = PACA
> + * PACA.exmc.CCR = guest CR
> + * PACA.exmc.R9 = guest R1
> + * PACA.exmc.R10 = guest R10
> + * PACA.exmc.R11 = guest R11
> + * PACA.exmc.R12 = guest R12
> + * PACA.exmc.R13 = guest R2
> + *
> + */
> +
> + /* Save registers */
> +
> + std r0, (PACA_EXMC+EX_SRR0)(r13)
> + std r9, (PACA_EXMC+EX_R3)(r13)
> + std r10, (PACA_EXMC+EX_LR)(r13)
> + std r11, (PACA_EXMC+EX_DAR)(r13)
> +
> + /*
> + * In order for us to easily get the last instruction,
> + * we got the #vmexit at, we exploit the fact that the
> + * virtual layout is still the same here, so we can just
> + * ld from the guest's PC address
> + */
> +
> + /* We only load the last instruction when it's safe */
> + cmpwi r12, BOOK3S_INTERRUPT_DATA_STORAGE
> + beq ld_last_inst
> + cmpwi r12, BOOK3S_INTERRUPT_PROGRAM
> + beq ld_last_inst
> +
> + b no_ld_last_inst
> +
> +ld_last_inst:
> + /* Save off the guest instruction we're at */
> + /* 1) enable paging for data */
> + mfmsr r9
> + ori r11, r9, MSR_DR /* Enable paging for data */
> + mtmsr r11
> + /* 2) fetch the instruction */
> + lwz r0, 0(r10)
> + /* 3) disable paging again */
> + mtmsr r9
> +
> +no_ld_last_inst:
> +
> + /* Restore bolted entries from the shadow and fix it along the way */
> +
> + /* We don't store anything in entry 0, so we don't need to take care of
that */
> + slbia
> + isync
> +
> +#if SLB_NUM_BOLTED == 3
> +
> + ld r11, PACA_SLBSHADOWPTR(r13)
> +
> + ld r10, 0x10(r11)
> + cmpdi r10, 0
> + beq slb_exit_skip_1
> + oris r10, r10, SLB_ESID_V@h
> + ld r9, 0x18(r11)
> + slbmte r9, r10
> + std r10, 0x10(r11)
> +slb_exit_skip_1:
> +
> + ld r10, 0x20(r11)
> + cmpdi r10, 0
> + beq slb_exit_skip_2
> + oris r10, r10, SLB_ESID_V@h
> + ld r9, 0x28(r11)
> + slbmte r9, r10
> + std r10, 0x20(r11)
> +slb_exit_skip_2:
> +
> + ld r10, 0x30(r11)
> + cmpdi r10, 0
> + beq slb_exit_skip_3
> + oris r10, r10, SLB_ESID_V@h
> + ld r9, 0x38(r11)
> + slbmte r9, r10
> + std r10, 0x30(r11)
> +slb_exit_skip_3:
Again, a macro here?
> +
> +#else
> +#error unknown number of bolted entries
> +#endif
> +
> +slb_do_exit:
> +
> + /* Restore registers */
> +
> + ld r11, (PACA_EXMC+EX_DAR)(r13)
> + ld r10, (PACA_EXMC+EX_LR)(r13)
> + ld r9, (PACA_EXMC+EX_R3)(r13)
> +
> + /* Save last inst */
> + stw r0, (PACA_EXMC+EX_LR)(r13)
> +
> + /* Save DAR and DSISR before going to paged mode */
> + mfdar r0
> + std r0, (PACA_EXMC+EX_DAR)(r13)
> + mfdsisr r0
> + stw r0, (PACA_EXMC+EX_DSISR)(r13)
> +
> + /* RFI into the highmem handler */
> + mfmsr r0
> + ori r0, r0, MSR_IR|MSR_DR|MSR_RI /* Enable paging */
> + mtsrr1 r0
> + ld r0, PACASAVEDMSR(r13) /* Highmem handler address */
> + mtsrr0 r0
> +
> + mfspr r0, SPRN_SPRG_SCRATCH0
> +
> + RFI
> +kvmppc_handler_trampoline_exit_end:
> +
> --
> 1.6.0.2
>
> _______________________________________________
> Linuxppc-dev mailing list
> Linuxppc-dev@lists.ozlabs.org
> https://lists.ozlabs.org/listinfo/linuxppc-dev
>
next prev parent reply other threads:[~2009-11-01 23:23 UTC|newest]
Thread overview: 49+ messages / expand[flat|nested] mbox.gz Atom feed top
2009-10-30 15:47 [PATCH 00/27] Add KVM support for Book3s_64 (PPC64) hosts v6 Alexander Graf
2009-10-30 15:47 ` [PATCH 02/27] Pass PVR in sregs Alexander Graf
2009-10-30 15:47 ` [PATCH 04/27] Add Book3s fields to vcpu structs Alexander Graf
2009-10-30 15:47 ` [PATCH 05/27] Add asm/kvm_book3s.h Alexander Graf
2009-10-30 15:47 ` [PATCH 06/27] Add Book3s_64 intercept helpers Alexander Graf
2009-10-30 15:47 ` [PATCH 11/27] Add book3s_64 Host MMU handling Alexander Graf
[not found] ` <1256917647-6200-12-git-send-email-agraf-l3A5Bk7waGM@public.gmane.org>
2009-11-01 23:39 ` Michael Neuling
2009-11-02 9:26 ` Alexander Graf
[not found] ` <1256917647-6200-1-git-send-email-agraf-l3A5Bk7waGM@public.gmane.org>
2009-10-30 15:47 ` [PATCH 01/27] Move dirty logging code to sub-arch Alexander Graf
2009-10-30 15:47 ` [PATCH 03/27] Add Book3s definitions Alexander Graf
2009-10-30 15:47 ` [PATCH 07/27] Add book3s_64 highmem asm code Alexander Graf
2009-10-30 15:47 ` [PATCH 08/27] Add SLB switching code for entry/exit Alexander Graf
2009-11-01 23:23 ` Michael Neuling [this message]
[not found] ` <6695.1257117827-/owAOxkjmzZAfugRpC6u6w@public.gmane.org>
2009-11-02 9:23 ` Alexander Graf
[not found] ` <00BF2D99-F2CE-4204-B4B4-0D113FD54CE6-l3A5Bk7waGM@public.gmane.org>
2009-11-02 9:39 ` Michael Neuling
2009-11-02 9:59 ` Alexander Graf
2009-10-30 15:47 ` [PATCH 09/27] Add interrupt handling code Alexander Graf
2009-10-30 15:47 ` [PATCH 10/27] Add book3s.c Alexander Graf
2009-10-30 15:47 ` [PATCH 12/27] Add book3s_64 guest MMU Alexander Graf
2009-10-30 15:47 ` [PATCH 13/27] Add book3s_32 " Alexander Graf
2009-10-30 15:47 ` [PATCH 14/27] Add book3s_64 specific opcode emulation Alexander Graf
2009-11-03 8:47 ` Segher Boessenkool
[not found] ` <A1CBD511-FF08-48BB-A8D6-9F66E20F770B-XVmvHMARGAS8U2dJNN8I7kB+6BGkLq7r@public.gmane.org>
2009-11-03 9:06 ` Alexander Graf
2009-11-03 21:38 ` Benjamin Herrenschmidt
2009-11-04 8:43 ` Arnd Bergmann
2009-11-04 8:47 ` Benjamin Herrenschmidt
2009-11-04 11:35 ` Alexander Graf
2009-11-05 0:53 ` Segher Boessenkool
2009-11-05 10:09 ` Alexander Graf
2009-10-30 15:47 ` [PATCH 15/27] Add mfdec emulation Alexander Graf
2009-10-30 15:47 ` [PATCH 16/27] Add desktop PowerPC specific emulation Alexander Graf
2009-10-30 15:47 ` [PATCH 19/27] Export symbols for KVM module Alexander Graf
[not found] ` <1256917647-6200-20-git-send-email-agraf-l3A5Bk7waGM@public.gmane.org>
2009-10-31 4:37 ` Stephen Rothwell
[not found] ` <20091031153719.10a4e61b.sfr-3FnU+UHB4dNDw9hX6IcOSA@public.gmane.org>
2009-10-31 12:02 ` Alexander Graf
2009-10-30 15:47 ` [PATCH 20/27] Split init_new_context and destroy_context Alexander Graf
2009-10-31 4:40 ` Stephen Rothwell
2009-10-31 21:20 ` Alexander Graf
2009-10-31 21:37 ` Benjamin Herrenschmidt
2009-10-30 15:47 ` [PATCH 21/27] Export KVM symbols for module Alexander Graf
2009-10-30 15:47 ` [PATCH 22/27] Add fields to PACA Alexander Graf
2009-10-30 15:47 ` [PATCH 27/27] Use hrtimers for the decrementer Alexander Graf
2009-11-05 6:03 ` [PATCH 00/27] Add KVM support for Book3s_64 (PPC64) hosts v6 Benjamin Herrenschmidt
2009-10-30 15:47 ` [PATCH 17/27] Make head_64.S aware of KVM real mode code Alexander Graf
2009-10-30 15:47 ` [PATCH 18/27] Add Book3s_64 offsets to asm-offsets.c Alexander Graf
2009-10-30 15:47 ` [PATCH 23/27] Export new PACA constants in asm-offsets Alexander Graf
2009-10-30 15:47 ` [PATCH 24/27] Include Book3s_64 target in buildsystem Alexander Graf
2009-10-30 15:47 ` [PATCH 25/27] Fix trace.h Alexander Graf
2009-10-30 15:47 ` [PATCH 26/27] Use Little Endian for Dirty Bitmap Alexander Graf
-- strict thread matches above, loose matches on Subject: below --
2009-10-21 15:03 [PATCH 00/27] Add KVM support for Book3s_64 (PPC64) hosts v5 Alexander Graf
2009-10-21 15:03 ` [PATCH 01/27] Move dirty logging code to sub-arch Alexander Graf
2009-10-21 15:03 ` [PATCH 02/27] Pass PVR in sregs Alexander Graf
2009-10-21 15:03 ` [PATCH 03/27] Add Book3s definitions Alexander Graf
2009-10-21 15:03 ` [PATCH 04/27] Add Book3s fields to vcpu structs Alexander Graf
2009-10-21 15:03 ` [PATCH 05/27] Add asm/kvm_book3s.h Alexander Graf
2009-10-21 15:03 ` [PATCH 06/27] Add Book3s_64 intercept helpers Alexander Graf
2009-10-21 15:03 ` [PATCH 07/27] Add book3s_64 highmem asm code Alexander Graf
[not found] ` <1256137413-15256-8-git-send-email-agraf-l3A5Bk7waGM@public.gmane.org>
2009-10-21 15:03 ` [PATCH 08/27] Add SLB switching code for entry/exit Alexander Graf
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=6695.1257117827@neuling.org \
--to=mikey@neuling.org \
--cc=agraf@suse.de \
--cc=arnd@arndb.de \
--cc=avi@redhat.com \
--cc=bphilips@suse.de \
--cc=hollisb@us.ibm.com \
--cc=kvm-ppc@vger.kernel.org \
--cc=kvm@vger.kernel.org \
--cc=kwolf@redhat.com \
--cc=linuxppc-dev@ozlabs.org \
--cc=mtosatti@redhat.com \
--cc=olof@lixom.net \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox