From: liweiwei <liweiwei@iscas.ac.cn>
To: Fei Wu <fei2.wu@intel.com>
Cc: liweiwei@iscas.ac.cn, Palmer Dabbelt <palmer@dabbelt.com>,
Alistair Francis <alistair.francis@wdc.com>,
Bin Meng <bin.meng@windriver.com>,
Daniel Henrique Barboza <dbarboza@ventanamicro.com>,
Liu Zhiwei <zhiwei_liu@linux.alibaba.com>,
"open list:RISC-V TCG CPUs" <qemu-riscv@nongnu.org>,
"open list:All patches CC here" <qemu-devel@nongnu.org>
Subject: Re: [PATCH v3] target/riscv: reduce overhead of MSTATUS_SUM change
Date: Wed, 22 Mar 2023 20:37:15 +0800 [thread overview]
Message-ID: <585bf39e-91e9-bd85-1fee-80ff0b4c0a8f@iscas.ac.cn> (raw)
In-Reply-To: <20230322121240.232303-1-fei2.wu@intel.com>
On 2023/3/22 20:12, Fei Wu wrote:
> Kernel needs to access user mode memory e.g. during syscalls, the window
> is usually opened up for a very limited time through MSTATUS.SUM, the
> overhead is too much if tlb_flush() gets called for every SUM change.
>
> This patch creates a separate MMU index for S+SUM, so that it's not
> necessary to flush tlb anymore when SUM changes. This is similar to how
> ARM handles Privileged Access Never (PAN).
>
> Result of 'pipe 10' from unixbench boosts from 223656 to 1705006. Many
> other syscalls benefit a lot from this too.
>
> Signed-off-by: Fei Wu <fei2.wu@intel.com>
> ---
> target/riscv/cpu-param.h | 2 +-
> target/riscv/cpu.h | 2 +-
> target/riscv/cpu_bits.h | 1 +
> target/riscv/cpu_helper.c | 11 +++++++++++
> target/riscv/csr.c | 2 +-
> 5 files changed, 15 insertions(+), 3 deletions(-)
>
> diff --git a/target/riscv/cpu-param.h b/target/riscv/cpu-param.h
> index ebaf26d26d..9e21b943f9 100644
> --- a/target/riscv/cpu-param.h
> +++ b/target/riscv/cpu-param.h
> @@ -27,6 +27,6 @@
> * - S mode HLV/HLVX/HSV 0b101
> * - M mode HLV/HLVX/HSV 0b111
> */
> -#define NB_MMU_MODES 8
> +#define NB_MMU_MODES 16
>
> #endif
> diff --git a/target/riscv/cpu.h b/target/riscv/cpu.h
> index 638e47c75a..ac8bee11a7 100644
> --- a/target/riscv/cpu.h
> +++ b/target/riscv/cpu.h
> @@ -624,7 +624,7 @@ target_ulong riscv_cpu_get_fflags(CPURISCVState *env);
> void riscv_cpu_set_fflags(CPURISCVState *env, target_ulong);
>
> #define TB_FLAGS_PRIV_MMU_MASK 3
> -#define TB_FLAGS_PRIV_HYP_ACCESS_MASK (1 << 2)
> +#define TB_FLAGS_PRIV_HYP_ACCESS_MASK (1 << 3)
> #define TB_FLAGS_MSTATUS_FS MSTATUS_FS
> #define TB_FLAGS_MSTATUS_VS MSTATUS_VS
>
> diff --git a/target/riscv/cpu_bits.h b/target/riscv/cpu_bits.h
> index fca7ef0cef..dd9e62b6e4 100644
> --- a/target/riscv/cpu_bits.h
> +++ b/target/riscv/cpu_bits.h
> @@ -606,6 +606,7 @@ typedef enum {
> #define PRV_S 1
> #define PRV_H 2 /* Reserved */
> #define PRV_M 3
> +#define MMUIdx_S_SUM 5
>
> /* Virtulisation Register Fields */
> #define VIRT_ONOFF 1
> diff --git a/target/riscv/cpu_helper.c b/target/riscv/cpu_helper.c
> index f88c503cf4..e52e9765d0 100644
> --- a/target/riscv/cpu_helper.c
> +++ b/target/riscv/cpu_helper.c
> @@ -36,6 +36,17 @@ int riscv_cpu_mmu_index(CPURISCVState *env, bool ifetch)
> #ifdef CONFIG_USER_ONLY
> return 0;
> #else
> + if (ifetch) {
> + return env->priv;
> + }
> +
> + int mode = env->priv;
> + if (mode == PRV_M && get_field(env->mstatus, MSTATUS_MPRV)) {
> + mode = get_field(env->mstatus, MSTATUS_MPP);
> + }
> + if (mode == PRV_S && get_field(env->mstatus, MSTATUS_SUM)) {
> + return MMUIdx_S_SUM;
> + }
> return env->priv;
If we return mode here, whether tlb needn't flush for changes to
MSTATUS_MPRV and MSTATUS_MPP?
Regards,
Weiwei Li
> #endif
> }
> diff --git a/target/riscv/csr.c b/target/riscv/csr.c
> index ab566639e5..eacc40e912 100644
> --- a/target/riscv/csr.c
> +++ b/target/riscv/csr.c
> @@ -1246,7 +1246,7 @@ static RISCVException write_mstatus(CPURISCVState *env, int csrno,
>
> /* flush tlb on mstatus fields that affect VM */
> if ((val ^ mstatus) & (MSTATUS_MXR | MSTATUS_MPP | MSTATUS_MPV |
> - MSTATUS_MPRV | MSTATUS_SUM)) {
> + MSTATUS_MPRV)) {
> tlb_flush(env_cpu(env));
> }
> mask = MSTATUS_SIE | MSTATUS_SPIE | MSTATUS_MIE | MSTATUS_MPIE |
next prev parent reply other threads:[~2023-03-22 12:37 UTC|newest]
Thread overview: 9+ messages / expand[flat|nested] mbox.gz Atom feed top
2023-03-22 12:12 [PATCH v3] target/riscv: reduce overhead of MSTATUS_SUM change Fei Wu
2023-03-22 12:37 ` liweiwei [this message]
2023-03-22 13:12 ` Wu, Fei
2023-03-22 13:19 ` Richard Henderson
2023-03-23 0:38 ` Wu, Fei
2023-03-23 1:26 ` Wu, Fei
2023-03-23 13:24 ` Wu, Fei
2023-03-23 16:41 ` Richard Henderson
2023-03-23 16:11 ` 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=585bf39e-91e9-bd85-1fee-80ff0b4c0a8f@iscas.ac.cn \
--to=liweiwei@iscas.ac.cn \
--cc=alistair.francis@wdc.com \
--cc=bin.meng@windriver.com \
--cc=dbarboza@ventanamicro.com \
--cc=fei2.wu@intel.com \
--cc=palmer@dabbelt.com \
--cc=qemu-devel@nongnu.org \
--cc=qemu-riscv@nongnu.org \
--cc=zhiwei_liu@linux.alibaba.com \
/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).