From: Charlie Jenkins <thecharlesjenkins@gmail.com>
To: Jesse Taube <jtaubepe@redhat.com>
Cc: linux-riscv@lists.infradead.org,
Albert Ou <aou@eecs.berkeley.edu>,
Alexandre Ghiti <alex@ghiti.fr>,
Daniel Lezcano <daniel.lezcano@linaro.org>,
linux-kernel@vger.kernel.org, Conor Dooley <conor@kernel.org>,
Samuel Holland <samuel.holland@sifive.com>,
Palmer Dabbelt <palmer@dabbelt.com>,
Jesse Taube <Mr.Bossman075@gmail.com>,
Jisheng Zhang <jszhang@kernel.org>,
Paul Walmsley <paul.walmsley@sifive.com>,
Thomas Gleixner <tglx@linutronix.de>
Subject: Re: [PATCH v7 1/2] clocksource/drivers/timer-clint: Add ZICNTR support
Date: Mon, 8 Jun 2026 06:54:48 -0700 [thread overview]
Message-ID: <aibJqFKQjCe30ejT@blinky> (raw)
In-Reply-To: <20260525182124.19402-1-jtaubepe@redhat.com>
On Mon, May 25, 2026 at 02:21:23PM -0400, Jesse Taube wrote:
> ZICNTR allows the CPU time to be read using CSRs, which used to be the
> default. If ZICNTR is supported prefer using CSRs over MMIO.
> This may also be faster as CSRs are not emulated in M-mode.
I think the rationale can be cleaned up a bit here. You say it "may" be
faster, if it's not guaranteed to be faster, why prefer it over mmio?
The patch d5be89a8d118 ("RISC-V: Resurrect the MMIO timer implementation
for M-mode systems") is the one that changed this behavior, can you
include a reference to that patch to point out when it was the default?
>
> Signed-off-by: Jesse Taube <jtaubepe@redhat.com>
> ---
> Treat this as a completely new patch, as it is mostly rewritten.
> Original:
> https://lore.kernel.org/all/20240410142347.964-3-jszhang@kernel.org/
> V3 -> V4:
> - Add riscv,csr-clint
> - Allow using of CSRs in S mode
> - Change if return else return to if return return
> - Change static_branch_likely to static_branch_unlikely
> - Fix 32-bit clint_get_cycles64 csr_available check being inverted
> - Fix is_c900_clint being uninitialized
> V4 -> V5:
> - Remove riscv,csr-clint
> - Remove riscv_csr_time_available
> - Replace all static_branch_unlikely(&riscv_csr_time_available) with
> riscv_has_extension_likely(RISCV_ISA_EXT_ZICNTR)
> V5 -> V6:
> - Add csr_clock_next_event
> - Add csr_get_cycles64 to sched_clock_register
> - Break get_cycles64 into csr_get_cycles64 and clint_get_cycles64
> - Change default behavior to use CSR time register
> V6 -> V7:
> - Split csr and c900 parts into different commits
>
> Posible future work:
> - Check for SSTC support and use stimercmp
> - SSTC depends on smode which cpufeature doesnt check for.
> ---
> arch/riscv/include/asm/timex.h | 12 ++++++-
> drivers/clocksource/timer-clint.c | 57 ++++++++++++++++++++++++++++---
> 2 files changed, 64 insertions(+), 5 deletions(-)
>
> diff --git a/arch/riscv/include/asm/timex.h b/arch/riscv/include/asm/timex.h
> index a06697846e69..d78075351bf4 100644
> --- a/arch/riscv/include/asm/timex.h
> +++ b/arch/riscv/include/asm/timex.h
> @@ -17,17 +17,26 @@ typedef unsigned long cycles_t;
> #ifdef CONFIG_64BIT
> static inline cycles_t get_cycles(void)
> {
> + if (riscv_has_extension_likely(RISCV_ISA_EXT_ZICNTR))
> + return csr_read(CSR_TIME);
> +
> return readq_relaxed(clint_time_val);
> }
> #else /* !CONFIG_64BIT */
> static inline u32 get_cycles(void)
> {
> + if (riscv_has_extension_likely(RISCV_ISA_EXT_ZICNTR))
> + return csr_read(CSR_TIME);
> +
> return readl_relaxed(((u32 *)clint_time_val));
> }
> #define get_cycles get_cycles
>
> static inline u32 get_cycles_hi(void)
> {
> + if (riscv_has_extension_likely(RISCV_ISA_EXT_ZICNTR))
> + return csr_read(CSR_TIMEH);
> +
> return readl_relaxed(((u32 *)clint_time_val) + 1);
> }
> #define get_cycles_hi get_cycles_hi
> @@ -40,7 +49,8 @@ static inline u32 get_cycles_hi(void)
> */
> static inline unsigned long random_get_entropy(void)
> {
> - if (unlikely(clint_time_val == NULL))
> + if (!riscv_has_extension_likely(RISCV_ISA_EXT_ZICNTR) &&
> + (unlikely(clint_time_val == NULL)))
> return random_get_entropy_fallback();
> return get_cycles();
> }
> diff --git a/drivers/clocksource/timer-clint.c b/drivers/clocksource/timer-clint.c
> index 0bdd9d7ec545..18f7bf7d26c6 100644
> --- a/drivers/clocksource/timer-clint.c
> +++ b/drivers/clocksource/timer-clint.c
> @@ -81,6 +81,12 @@ static u64 notrace clint_get_cycles64(void)
> {
> return clint_get_cycles();
> }
> +
> +static u64 notrace csr_get_cycles64(void)
> +{
> + return csr_read(CSR_TIME);
> +}
> +
> #else /* CONFIG_64BIT */
> static u64 notrace clint_get_cycles64(void)
> {
> @@ -93,8 +99,26 @@ static u64 notrace clint_get_cycles64(void)
>
> return ((u64)hi << 32) | lo;
> }
> +
> +static u64 notrace csr_get_cycles64(void)
> +{
> + u32 hi, lo;
> +
> + do {
> + hi = csr_read(CSR_TIMEH);
> + lo = csr_read(CSR_TIME);
> + } while (hi != csr_read(CSR_TIMEH));
> +
> + return ((u64)hi << 32) | lo;
> +}
> +
> #endif /* CONFIG_64BIT */
>
> +static u64 csr_rdtime(struct clocksource *cs)
> +{
> + return csr_get_cycles64();
> +}
> +
> static u64 clint_rdtime(struct clocksource *cs)
> {
> return clint_get_cycles64();
> @@ -105,9 +129,20 @@ static struct clocksource clint_clocksource = {
> .rating = 300,
> .mask = CLOCKSOURCE_MASK(64),
> .flags = CLOCK_SOURCE_IS_CONTINUOUS,
> - .read = clint_rdtime,
> + .read = csr_rdtime,
> };
>
> +static int csr_clock_next_event(unsigned long delta,
> + struct clock_event_device *ce)
> +{
> + void __iomem *r = clint_timer_cmp +
> + cpuid_to_hartid_map(smp_processor_id());
> +
> + csr_set(CSR_IE, IE_TIE);
> + writeq_relaxed(csr_get_cycles64() + delta, r);
> + return 0;
> +}
> +
> static int clint_clock_next_event(unsigned long delta,
> struct clock_event_device *ce)
> {
> @@ -123,13 +158,16 @@ static DEFINE_PER_CPU(struct clock_event_device, clint_clock_event) = {
> .name = "clint_clockevent",
> .features = CLOCK_EVT_FEAT_ONESHOT,
> .rating = 100,
> - .set_next_event = clint_clock_next_event,
> + .set_next_event = csr_clock_next_event,
> };
>
> static int clint_timer_starting_cpu(unsigned int cpu)
> {
> struct clock_event_device *ce = per_cpu_ptr(&clint_clock_event, cpu);
>
> + if (!riscv_has_extension_likely(RISCV_ISA_EXT_ZICNTR))
> + ce->set_next_event = clint_clock_next_event;
> +
> ce->cpumask = cpumask_of(cpu);
> clockevents_config_and_register(ce, clint_timer_freq, 100, ULONG_MAX);
>
> @@ -161,7 +199,7 @@ static irqreturn_t clint_timer_interrupt(int irq, void *dev_id)
> return IRQ_HANDLED;
> }
>
> -static int __init clint_timer_init_dt(struct device_node *np)
> +static int __init clint_timer_init(struct device_node *np)
> {
> int rc;
> u32 i, nr_irqs;
> @@ -227,13 +265,19 @@ static int __init clint_timer_init_dt(struct device_node *np)
>
> pr_info("%pOFP: timer running at %ld Hz\n", np, clint_timer_freq);
>
> + if (!riscv_has_extension_likely(RISCV_ISA_EXT_ZICNTR))
> + clint_clocksource.read = clint_rdtime;
> +
> rc = clocksource_register_hz(&clint_clocksource, clint_timer_freq);
> if (rc) {
> pr_err("%pOFP: clocksource register failed [%d]\n", np, rc);
> goto fail_iounmap;
> }
>
> - sched_clock_register(clint_get_cycles64, 64, clint_timer_freq);
> + if (riscv_has_extension_likely(RISCV_ISA_EXT_ZICNTR))
> + sched_clock_register(csr_get_cycles64, 64, clint_timer_freq);
> + else
> + sched_clock_register(clint_get_cycles64, 64, clint_timer_freq);
>
> rc = request_percpu_irq(clint_timer_irq, clint_timer_interrupt,
> "clint-timer", &clint_clock_event);
> @@ -273,5 +317,10 @@ static int __init clint_timer_init_dt(struct device_node *np)
> return rc;
> }
>
> +static int __init clint_timer_init_dt(struct device_node *np)
> +{
> + return clint_timer_init(np);
> +}
> +
> TIMER_OF_DECLARE(clint_timer, "riscv,clint0", clint_timer_init_dt);
> TIMER_OF_DECLARE(clint_timer1, "sifive,clint0", clint_timer_init_dt);
> --
> 2.54.0
>
>
> _______________________________________________
> linux-riscv mailing list
> linux-riscv@lists.infradead.org
> http://lists.infradead.org/mailman/listinfo/linux-riscv
>
_______________________________________________
linux-riscv mailing list
linux-riscv@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-riscv
prev parent reply other threads:[~2026-06-08 13:55 UTC|newest]
Thread overview: 3+ messages / expand[flat|nested] mbox.gz Atom feed top
2026-05-25 18:21 [PATCH v7 1/2] clocksource/drivers/timer-clint: Add ZICNTR support Jesse Taube
2026-05-25 18:21 ` [PATCH v7 2/2] clocksource/drivers/timer-clint: Add T-Head C9xx clint Jesse Taube
2026-06-08 13:54 ` Charlie Jenkins [this message]
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=aibJqFKQjCe30ejT@blinky \
--to=thecharlesjenkins@gmail.com \
--cc=Mr.Bossman075@gmail.com \
--cc=alex@ghiti.fr \
--cc=aou@eecs.berkeley.edu \
--cc=conor@kernel.org \
--cc=daniel.lezcano@linaro.org \
--cc=jszhang@kernel.org \
--cc=jtaubepe@redhat.com \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-riscv@lists.infradead.org \
--cc=palmer@dabbelt.com \
--cc=paul.walmsley@sifive.com \
--cc=samuel.holland@sifive.com \
--cc=tglx@linutronix.de \
/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