From: "Clément Léger" <cleger@rivosinc.com>
To: Jesse Taube <jesse@rivosinc.com>,
linux-riscv@lists.infradead.org, linux-kernel@vger.kernel.org
Cc: Paul Walmsley <paul.walmsley@sifive.com>,
Palmer Dabbelt <palmer@dabbelt.com>,
Albert Ou <aou@eecs.berkeley.edu>,
Alexandre Ghiti <alex@ghiti.fr>, Oleg Nesterov <oleg@redhat.com>,
Himanshu Chauhan <hchauhan@ventanamicro.com>,
Charlie Jenkins <charlie@rivosinc.com>,
Samuel Holland <samuel.holland@sifive.com>,
Deepak Gupta <debug@rivosinc.com>,
Andrew Jones <ajones@ventanamicro.com>,
Atish Patra <atishp@rivosinc.com>,
Anup Patel <apatel@ventanamicro.com>,
Mayuresh Chitale <mchitale@ventanamicro.com>,
Conor Dooley <conor.dooley@microchip.com>,
WangYuli <wangyuli@uniontech.com>,
Huacai Chen <chenhuacai@kernel.org>,
Nam Cao <namcao@linutronix.de>,
Andrew Morton <akpm@linux-foundation.org>,
"Mike Rapoport (Microsoft)" <rppt@kernel.org>,
Luis Chamberlain <mcgrof@kernel.org>,
Yunhui Cui <cuiyunhui@bytedance.com>,
Joel Granados <joel.granados@kernel.org>,
Celeste Liu <coelacanthushex@gmail.com>,
Evan Green <evan@rivosinc.com>,
Nylon Chen <nylon.chen@sifive.com>
Subject: Re: [RFC PATCH 5/6] riscv: hw_breakpoint: Use icount for single stepping
Date: Mon, 4 Aug 2025 10:15:22 +0200 [thread overview]
Message-ID: <b37f6f25-7569-4cae-b33d-61ed14b19611@rivosinc.com> (raw)
In-Reply-To: <20250722173829.984082-6-jesse@rivosinc.com>
On 22/07/2025 19:38, Jesse Taube wrote:
> The Sdtrig RISC-V ISA extension does not have a resume flag for
> returning to and executing the instruction at the breakpoint.
> To avoid skipping the instruction or looping, it is necessary to remove
> the hardware breakpoint and single step. Use the icount feature of
> Sdtrig to accomplish this. Use icount as default with an option to allow
> software-based single stepping when hardware or SBI does not have
> icount functionality, as it may cause unwanted side effects when reading
> the instruction from memory.
>
> Signed-off-by: Jesse Taube <jesse@rivosinc.com>
> ---
> arch/riscv/Kconfig | 11 +++++
> arch/riscv/kernel/hw_breakpoint.c | 81 +++++++++++++++++++++++++------
> 2 files changed, 76 insertions(+), 16 deletions(-)
>
> diff --git a/arch/riscv/Kconfig b/arch/riscv/Kconfig
> index 95d3047cab10..bbde5e118470 100644
> --- a/arch/riscv/Kconfig
> +++ b/arch/riscv/Kconfig
> @@ -1105,6 +1105,17 @@ config PARAVIRT_TIME_ACCOUNTING
>
> If in doubt, say N here.
>
> +config HW_BREAKPOINT_COMPUTE_STEP
> + bool "Allow computing hardware breakpoint step address"
> + default n
> + depends on HAVE_HW_BREAKPOINT
> + help
> + Select this option if hardware breakpoints are desired, but
> + hardware or SBI does not have icount functionality. This may cause
> + unwanted side affects when reading the instruction from memory.
Hi Jesse,
Typo: s/affects/effects
Thanks,
Clément
> +
> + If unsure, say N.
> +
> config RELOCATABLE
> bool "Build a relocatable kernel"
> depends on !XIP_KERNEL
> diff --git a/arch/riscv/kernel/hw_breakpoint.c b/arch/riscv/kernel/hw_breakpoint.c
> index 9e3a3b82d300..437fd82b9590 100644
> --- a/arch/riscv/kernel/hw_breakpoint.c
> +++ b/arch/riscv/kernel/hw_breakpoint.c
> @@ -20,6 +20,7 @@
> #define DBTR_TDATA1_DMODE BIT_UL(__riscv_xlen - 5)
>
> #define DBTR_TDATA1_TYPE_MCONTROL (2UL << DBTR_TDATA1_TYPE_SHIFT)
> +#define DBTR_TDATA1_TYPE_ICOUNT (3UL << DBTR_TDATA1_TYPE_SHIFT)
> #define DBTR_TDATA1_TYPE_MCONTROL6 (6UL << DBTR_TDATA1_TYPE_SHIFT)
>
> #define DBTR_TDATA1_MCONTROL6_LOAD BIT(0)
> @@ -55,6 +56,14 @@
> #define DBTR_TDATA1_MCONTROL_SIZELO_64BIT 1
> #define DBTR_TDATA1_MCONTROL_SIZEHI_64BIT 1
>
> +#define DBTR_TDATA1_ICOUNT_U BIT(6)
> +#define DBTR_TDATA1_ICOUNT_S BIT(7)
> +#define DBTR_TDATA1_ICOUNT_PENDING BIT(8)
> +#define DBTR_TDATA1_ICOUNT_M BIT(9)
> +#define DBTR_TDATA1_ICOUNT_COUNT_FIELD GENMASK(23, 10)
> +#define DBTR_TDATA1_ICOUNT_VU BIT(25)
> +#define DBTR_TDATA1_ICOUNT_VS BIT(26)
> +
> /* Registered per-cpu bp/wp */
> static DEFINE_PER_CPU(struct perf_event *, pcpu_hw_bp_events[RV_MAX_TRIGGERS]);
> static DEFINE_PER_CPU(unsigned long, ecall_lock_flags);
> @@ -65,6 +74,7 @@ static DEFINE_PER_CPU(union sbi_dbtr_shmem_entry, sbi_dbtr_shmem);
>
> /* number of debug triggers on this cpu . */
> static int dbtr_total_num __ro_after_init;
> +static bool have_icount __ro_after_init;
> static unsigned long dbtr_type __ro_after_init;
> static unsigned long dbtr_init __ro_after_init;
>
> @@ -168,6 +178,7 @@ static int arch_smp_teardown_sbi_shmem(unsigned int cpu)
> static void init_sbi_dbtr(void)
> {
> struct sbiret ret;
> + unsigned long dbtr_count = 0;
>
> /*
> * Called by hw_breakpoint_slots and arch_hw_breakpoint_init.
> @@ -182,6 +193,25 @@ static void init_sbi_dbtr(void)
> return;
> }
>
> + ret = sbi_ecall(SBI_EXT_DBTR, SBI_EXT_DBTR_NUM_TRIGGERS,
> + DBTR_TDATA1_TYPE_ICOUNT, 0, 0, 0, 0, 0);
> + if (ret.error) {
> + pr_warn("%s: failed to detect icount triggers. error: %ld.\n",
> + __func__, ret.error);
> + } else if (!ret.value) {
> + if (IS_ENABLED(CONFIG_HW_BREAKPOINT_COMPUTE_STEP)) {
> + pr_warn("%s: No icount triggers available. "
> + "Falling-back to computing single step address.\n", __func__);
> + } else {
> + pr_err("%s: No icount triggers available.\n", __func__);
> + dbtr_total_num = 0;
> + return;
> + }
> + } else {
> + dbtr_count = ret.value;
> + have_icount = true;
> + }
> +
> ret = sbi_ecall(SBI_EXT_DBTR, SBI_EXT_DBTR_NUM_TRIGGERS,
> DBTR_TDATA1_TYPE_MCONTROL6, 0, 0, 0, 0, 0);
> if (ret.error) {
> @@ -190,7 +220,7 @@ static void init_sbi_dbtr(void)
> } else if (!ret.value) {
> pr_warn("%s: No mcontrol6 triggers available.\n", __func__);
> } else {
> - dbtr_total_num = ret.value;
> + dbtr_total_num = min_not_zero((unsigned long)ret.value, dbtr_count);
> dbtr_type = DBTR_TDATA1_TYPE_MCONTROL6;
> return;
> }
> @@ -205,7 +235,7 @@ static void init_sbi_dbtr(void)
> pr_err("%s: No mcontrol triggers available.\n", __func__);
> dbtr_total_num = 0;
> } else {
> - dbtr_total_num = ret.value;
> + dbtr_total_num = min_not_zero((unsigned long)ret.value, dbtr_count);
> dbtr_type = DBTR_TDATA1_TYPE_MCONTROL;
> }
> }
> @@ -344,6 +374,21 @@ static int rv_init_mcontrol6_trigger(const struct perf_event_attr *attr,
> return 0;
> }
>
> +static int rv_init_icount_trigger(struct arch_hw_breakpoint *hw)
> +{
> + unsigned long tdata1 = DBTR_TDATA1_TYPE_ICOUNT;
> +
> + /* Step one instruction */
> + tdata1 |= FIELD_PREP(DBTR_TDATA1_ICOUNT_COUNT_FIELD, 1);
> +
> + tdata1 |= DBTR_TDATA1_ICOUNT_U;
> +
> + hw->tdata1 = tdata1;
> + hw->tdata2 = 0;
> +
> + return 0;
> +}
> +
> int hw_breakpoint_arch_parse(struct perf_event *bp,
> const struct perf_event_attr *attr,
> struct arch_hw_breakpoint *hw)
> @@ -389,24 +434,28 @@ static int setup_singlestep(struct perf_event *event, struct pt_regs *regs)
> /* Remove breakpoint even if return error as not to loop */
> arch_uninstall_hw_breakpoint(event);
>
> - ret = get_insn(regs, regs->epc, &insn);
> - if (ret < 0)
> - return ret;
> + if (have_icount) {
> + rv_init_icount_trigger(bp);
> + } else {
> + ret = get_insn(regs, regs->epc, &insn);
> + if (ret < 0)
> + return ret;
>
> - next_addr = get_step_address(regs, insn);
> + next_addr = get_step_address(regs, insn);
>
> - ret = get_insn(regs, next_addr, &insn);
> - if (ret < 0)
> - return ret;
> + ret = get_insn(regs, next_addr, &insn);
> + if (ret < 0)
> + return ret;
>
> - bp_insn.bp_type = HW_BREAKPOINT_X;
> - bp_insn.bp_addr = next_addr;
> - /* Get the size of the intruction */
> - bp_insn.bp_len = GET_INSN_LENGTH(insn);
> + bp_insn.bp_type = HW_BREAKPOINT_X;
> + bp_insn.bp_addr = next_addr;
> + /* Get the size of the intruction */
> + bp_insn.bp_len = GET_INSN_LENGTH(insn);
>
> - ret = hw_breakpoint_arch_parse(NULL, &bp_insn, bp);
> - if (ret)
> - return ret;
> + ret = hw_breakpoint_arch_parse(NULL, &bp_insn, bp);
> + if (ret)
> + return ret;
> + }
>
> ret = arch_install_hw_breakpoint(event);
> if (ret)
next prev parent reply other threads:[~2025-08-04 8:15 UTC|newest]
Thread overview: 18+ messages / expand[flat|nested] mbox.gz Atom feed top
2025-07-22 17:38 [RFC PATCH 0/6] riscv: add initial support for hardware breakpoints Jesse Taube
2025-07-22 17:38 ` [RFC PATCH 1/6] riscv: Add insn.c, consolidate instruction decoding Jesse Taube
2025-07-22 17:38 ` [RFC PATCH 2/6] riscv: Add SBI debug trigger extension and function ids Jesse Taube
2025-07-22 17:38 ` [RFC PATCH 3/6] riscv: insn: __read_insn use copy_from_X_nofault Jesse Taube
2025-08-04 7:44 ` Clément Léger
2025-07-22 17:38 ` [RFC PATCH 4/6] riscv: Introduce support for hardware break/watchpoints Jesse Taube
2025-07-23 2:49 ` Deepak Gupta
2025-07-23 17:02 ` Jesse Taube
2025-07-23 17:25 ` Deepak Gupta
2025-08-04 8:01 ` Clément Léger
2025-08-04 15:03 ` Jesse Taube
2025-07-22 17:38 ` [RFC PATCH 5/6] riscv: hw_breakpoint: Use icount for single stepping Jesse Taube
2025-08-04 8:15 ` Clément Léger [this message]
2025-07-22 17:38 ` [RFC PATCH 6/6] riscv: ptrace: Add hw breakpoint support Jesse Taube
2025-07-23 4:18 ` Deepak Gupta
2025-07-23 16:55 ` Jesse Taube
2025-07-23 17:23 ` Deepak Gupta
2025-08-04 8:16 ` Clément Léger
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=b37f6f25-7569-4cae-b33d-61ed14b19611@rivosinc.com \
--to=cleger@rivosinc.com \
--cc=ajones@ventanamicro.com \
--cc=akpm@linux-foundation.org \
--cc=alex@ghiti.fr \
--cc=aou@eecs.berkeley.edu \
--cc=apatel@ventanamicro.com \
--cc=atishp@rivosinc.com \
--cc=charlie@rivosinc.com \
--cc=chenhuacai@kernel.org \
--cc=coelacanthushex@gmail.com \
--cc=conor.dooley@microchip.com \
--cc=cuiyunhui@bytedance.com \
--cc=debug@rivosinc.com \
--cc=evan@rivosinc.com \
--cc=hchauhan@ventanamicro.com \
--cc=jesse@rivosinc.com \
--cc=joel.granados@kernel.org \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-riscv@lists.infradead.org \
--cc=mcgrof@kernel.org \
--cc=mchitale@ventanamicro.com \
--cc=namcao@linutronix.de \
--cc=nylon.chen@sifive.com \
--cc=oleg@redhat.com \
--cc=palmer@dabbelt.com \
--cc=paul.walmsley@sifive.com \
--cc=rppt@kernel.org \
--cc=samuel.holland@sifive.com \
--cc=wangyuli@uniontech.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
all inboxes | Powered by JetHome®