From: Peter Maydell <peter.maydell@linaro.org>
To: Richard Henderson <richard.henderson@linaro.org>
Cc: qemu-arm <qemu-arm@nongnu.org>, QEMU Developers <qemu-devel@nongnu.org>
Subject: Re: [PATCH v5 21/22] target/arm: Add mmu indexes for tag memory
Date: Fri, 6 Dec 2019 11:46:02 +0000 [thread overview]
Message-ID: <CAFEAcA8aJZRMFEpKaUN_W4U5LXoWvvzYUe0FQYwTd=RiFHeFEQ@mail.gmail.com> (raw)
In-Reply-To: <20191011134744.2477-22-richard.henderson@linaro.org>
On Fri, 11 Oct 2019 at 14:50, Richard Henderson
<richard.henderson@linaro.org> wrote:
>
> The process by which one goes from an address space plus physical
> address to a host pointer is complex. It is easiest to reuse the
> mechanism already present within cputlb, and letting that cache
> the results.
>
> Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
> ---
> target/arm/cpu-param.h | 2 +-
> target/arm/cpu.h | 12 +++++++++---
> target/arm/internals.h | 2 ++
> target/arm/helper.c | 25 +++++++++++++++++++++++--
> 4 files changed, 35 insertions(+), 6 deletions(-)
>
> diff --git a/target/arm/cpu-param.h b/target/arm/cpu-param.h
> index 6e6948e960..18ac562346 100644
> --- a/target/arm/cpu-param.h
> +++ b/target/arm/cpu-param.h
> @@ -29,6 +29,6 @@
> # define TARGET_PAGE_BITS_MIN 10
> #endif
>
> -#define NB_MMU_MODES 8
> +#define NB_MMU_MODES 9
>
> #endif
> diff --git a/target/arm/cpu.h b/target/arm/cpu.h
> index faca43ea78..c3609ef9d5 100644
> --- a/target/arm/cpu.h
> +++ b/target/arm/cpu.h
> @@ -2854,8 +2854,8 @@ static inline bool arm_excp_unmasked(CPUState *cs, unsigned int excp_idx,
> #define ARM_MMU_IDX_M_NEGPRI 0x2
> #define ARM_MMU_IDX_M_S 0x4
>
> -#define ARM_MMU_IDX_TYPE_MASK (~0x7)
> -#define ARM_MMU_IDX_COREIDX_MASK 0x7
> +#define ARM_MMU_IDX_TYPE_MASK (~0xf)
> +#define ARM_MMU_IDX_COREIDX_MASK 0xf
>
> typedef enum ARMMMUIdx {
> ARMMMUIdx_S12NSE0 = 0 | ARM_MMU_IDX_A,
> @@ -2865,6 +2865,9 @@ typedef enum ARMMMUIdx {
> ARMMMUIdx_S1SE0 = 4 | ARM_MMU_IDX_A,
> ARMMMUIdx_S1SE1 = 5 | ARM_MMU_IDX_A,
> ARMMMUIdx_S2NS = 6 | ARM_MMU_IDX_A,
> + ARMMMUIdx_TagNS = 7 | ARM_MMU_IDX_A,
> + ARMMMUIdx_TagS = 8 | ARM_MMU_IDX_A,
> +
> ARMMMUIdx_MUser = 0 | ARM_MMU_IDX_M,
> ARMMMUIdx_MPriv = 1 | ARM_MMU_IDX_M,
> ARMMMUIdx_MUserNegPri = 2 | ARM_MMU_IDX_M,
> @@ -2891,6 +2894,8 @@ typedef enum ARMMMUIdxBit {
> ARMMMUIdxBit_S1SE0 = 1 << 4,
> ARMMMUIdxBit_S1SE1 = 1 << 5,
> ARMMMUIdxBit_S2NS = 1 << 6,
> + ARMMMUIdxBit_TagNS = 1 << 7,
> + ARMMMUIdxBit_TagS = 1 << 8,
> ARMMMUIdxBit_MUser = 1 << 0,
> ARMMMUIdxBit_MPriv = 1 << 1,
> ARMMMUIdxBit_MUserNegPri = 1 << 2,
> @@ -3254,7 +3259,8 @@ enum {
> /* Return the address space index to use for a memory access */
> static inline int arm_asidx_from_attrs(CPUState *cs, MemTxAttrs attrs)
> {
> - return attrs.secure ? ARMASIdx_S : ARMASIdx_NS;
> + return ((attrs.target_tlb_bit2 ? ARMASIdx_TagNS : ARMASIdx_NS)
> + + attrs.secure);
If you want to do the "just add attrs.secure" can we have
a build-time assert that ARMASIdx_S is ARMASIdx_NS + 1, and
ditto for TagNS/TagS, please? It seems like the kind of thing
that will catch us out later on.
> }
> if (env->cp15.hcr_el2 & HCR_TGE) {
> @@ -10662,7 +10671,9 @@ bool get_phys_addr(CPUARMState *env, target_ulong address,
> target_ulong *page_size,
> ARMMMUFaultInfo *fi, ARMCacheAttrs *cacheattrs)
> {
> - if (mmu_idx == ARMMMUIdx_S12NSE0 || mmu_idx == ARMMMUIdx_S12NSE1) {
> + switch (mmu_idx) {
> + case ARMMMUIdx_S12NSE0:
> + case ARMMMUIdx_S12NSE1:
> /* Call ourselves recursively to do the stage 1 and then stage 2
> * translations.
> */
> @@ -10713,6 +10724,16 @@ bool get_phys_addr(CPUARMState *env, target_ulong address,
> */
> mmu_idx = stage_1_mmu_idx(mmu_idx);
> }
> + break;
> +
> + case ARMMMUIdx_TagS:
> + case ARMMMUIdx_TagNS:
> + /* Indicate tag memory to arm_asidx_from_attrs. */
> + attrs->target_tlb_bit2 = true;
> + break;
So here we fall through to the "handle a stage 1 lookup" code, which:
* sets attrs->secure
* sets attrs->user (always false, so we could have left it alone)
* skips the FCSE handling (as we're v8)
* skips the PMSA handling
* hits the regime_translation_disabled() check, which fills in
*phys_ptr, *prot and *page_size and returns
Maybe it would be clearer if this case here just did all of that:
case ARMMMUIdx_TagS:
attrs->secure = true;
/* fall through */
case ARMMMUIdx_TagNS:
/* Indicate tag memory to arm_asidx_from_attrs. */
attrs->target_tlb_bit2 = true;
*phys_ptr = address;
*prot = PAGE_READ | PAGE_WRITE | PAGE_EXEC;
*page_size = TARGET_PAGE_SIZE;
return 0;
Did I miss anything out?
Or are we going to want more things which are common between
the stage 1 codepath and the "just a tag ram access" case
in future?
thanks
-- PMM
next prev parent reply other threads:[~2019-12-06 15:38 UTC|newest]
Thread overview: 58+ messages / expand[flat|nested] mbox.gz Atom feed top
2019-10-11 13:47 [PATCH v5 00/22] target/arm: Implement ARMv8.5-MemTag, system mode Richard Henderson
2019-10-11 13:47 ` [PATCH v5 01/22] target/arm: Add MTE_ACTIVE to tb_flags Richard Henderson
2019-10-11 13:47 ` [PATCH v5 02/22] target/arm: Add regime_has_2_ranges Richard Henderson
2019-12-03 11:01 ` Peter Maydell
2019-12-03 15:09 ` Richard Henderson
2019-10-11 13:47 ` [PATCH v5 03/22] target/arm: Add MTE system registers Richard Henderson
2019-12-03 11:48 ` Peter Maydell
2019-12-06 14:47 ` Richard Henderson
2019-10-11 13:47 ` [PATCH v5 04/22] target/arm: Add helper_mte_check{1,2,3} Richard Henderson
2019-12-03 13:42 ` Peter Maydell
2019-12-03 16:06 ` Richard Henderson
2019-12-03 16:26 ` Peter Maydell
2019-12-03 16:14 ` Richard Henderson
2019-10-11 13:47 ` [PATCH v5 05/22] target/arm: Suppress tag check for sp+offset Richard Henderson
2019-12-03 14:07 ` Peter Maydell
2020-02-17 21:32 ` Richard Henderson
2019-10-11 13:47 ` [PATCH v5 06/22] target/arm: Implement the IRG instruction Richard Henderson
2019-12-03 14:26 ` Peter Maydell
2019-10-11 13:47 ` [PATCH v5 07/22] target/arm: Implement ADDG, SUBG instructions Richard Henderson
2019-10-11 13:47 ` [PATCH v5 08/22] target/arm: Implement the GMI instruction Richard Henderson
2019-10-11 13:47 ` [PATCH v5 09/22] target/arm: Implement the SUBP instruction Richard Henderson
2019-10-11 13:47 ` [PATCH v5 10/22] target/arm: Define arm_cpu_do_unaligned_access for CONFIG_USER_ONLY Richard Henderson
2019-12-05 16:12 ` Peter Maydell
2020-02-17 22:56 ` Richard Henderson
2019-10-11 13:47 ` [PATCH v5 11/22] target/arm: Implement LDG, STG, ST2G instructions Richard Henderson
2019-12-05 17:07 ` Peter Maydell
2019-10-11 13:47 ` [PATCH v5 12/22] target/arm: Implement the STGP instruction Richard Henderson
2019-12-05 17:15 ` Peter Maydell
2019-10-11 13:47 ` [PATCH v5 13/22] target/arm: Implement the LDGM and STGM instructions Richard Henderson
2019-12-05 17:42 ` Peter Maydell
2019-10-11 13:47 ` [PATCH v5 14/22] target/arm: Implement the access tag cache flushes Richard Henderson
2019-12-05 17:49 ` Peter Maydell
2019-10-11 13:47 ` [PATCH v5 15/22] target/arm: Clean address for DC ZVA Richard Henderson
2019-12-05 17:54 ` Peter Maydell
2019-12-05 18:58 ` Peter Maydell
2020-02-18 0:50 ` Richard Henderson
2020-02-18 11:10 ` Peter Maydell
2019-10-11 13:47 ` [PATCH v5 16/22] target/arm: Implement data cache set allocation tags Richard Henderson
2019-12-05 18:17 ` Peter Maydell
2020-02-18 1:19 ` Richard Henderson
2019-10-11 13:47 ` [PATCH v5 17/22] target/arm: Set PSTATE.TCO on exception entry Richard Henderson
2019-10-11 13:47 ` [PATCH v5 18/22] target/arm: Enable MTE Richard Henderson
2019-12-05 18:23 ` Peter Maydell
2019-10-11 13:47 ` [PATCH v5 19/22] target/arm: Cache the Tagged bit for a page in MemTxAttrs Richard Henderson
2019-12-05 18:32 ` Peter Maydell
2019-10-11 13:47 ` [PATCH v5 20/22] target/arm: Create tagged ram when MTE is enabled Richard Henderson
2019-12-05 18:40 ` Peter Maydell
2019-12-05 19:24 ` Richard Henderson
2019-12-06 9:51 ` Peter Maydell
2019-10-11 13:47 ` [PATCH v5 21/22] target/arm: Add mmu indexes for tag memory Richard Henderson
2019-12-06 11:46 ` Peter Maydell [this message]
2019-12-06 14:03 ` Richard Henderson
2019-10-11 13:47 ` [PATCH v5 22/22] target/arm: Add allocation tag storage for system mode Richard Henderson
2019-12-06 13:02 ` Peter Maydell
2019-12-06 14:14 ` Richard Henderson
2019-10-11 19:32 ` [PATCH v5 00/22] target/arm: Implement ARMv8.5-MemTag, " no-reply
2019-10-15 20:39 ` Evgenii Stepanov
2019-10-15 22:04 ` Richard Henderson
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='CAFEAcA8aJZRMFEpKaUN_W4U5LXoWvvzYUe0FQYwTd=RiFHeFEQ@mail.gmail.com' \
--to=peter.maydell@linaro.org \
--cc=qemu-arm@nongnu.org \
--cc=qemu-devel@nongnu.org \
--cc=richard.henderson@linaro.org \
/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;
as well as URLs for NNTP newsgroup(s).