From: Nikhil Gupta <[email protected]>
1588 driver loses about 1us in adjtime operation at PTP slave
This is because adjtime operation uses a slow non-atomic tmr_cnt_read()
followed by tmr_cnt_write() operation.
In the above sequence, since the timer counter operation keeps
incrementing, it leads to latency. The tmr_offset register
(which is added to TMR_CNT_H/L register giving the current time)
must be programmed with the delta nanoseconds.
Signed-off-by: Nikhil Gupta <[email protected]>
---
v1->v2: prevent TMR_OFF adjustment in case of eTSEC
drivers/ptp/ptp_qoriq.c | 50 ++++++++++++++++++++++++++++++-----
include/linux/fsl/ptp_qoriq.h | 1 +
2 files changed, 44 insertions(+), 7 deletions(-)
diff --git a/drivers/ptp/ptp_qoriq.c b/drivers/ptp/ptp_qoriq.c
index 08f4cf0ad9e3..61530167efe4 100644
--- a/drivers/ptp/ptp_qoriq.c
+++ b/drivers/ptp/ptp_qoriq.c
@@ -48,6 +48,29 @@ static void tmr_cnt_write(struct ptp_qoriq *ptp_qoriq, u64 ns)
ptp_qoriq->write(®s->ctrl_regs->tmr_cnt_h, hi);
}
+static u64 tmr_offset_read(struct ptp_qoriq *ptp_qoriq)
+{
+ struct ptp_qoriq_registers *regs = &ptp_qoriq->regs;
+ u32 lo, hi;
+ u64 ns;
+
+ lo = ptp_qoriq->read(®s->ctrl_regs->tmroff_l);
+ hi = ptp_qoriq->read(®s->ctrl_regs->tmroff_h);
+ ns = ((u64) hi) << 32;
+ ns |= lo;
+ return ns;
+}
+
+static void tmr_offset_write(struct ptp_qoriq *ptp_qoriq, u64 delta_ns)
+{
+ struct ptp_qoriq_registers *regs = &ptp_qoriq->regs;
+ u32 lo = delta_ns & 0xffffffff;
+ u32 hi = delta_ns >> 32;
+
+ ptp_qoriq->write(®s->ctrl_regs->tmroff_l, lo);
+ ptp_qoriq->write(®s->ctrl_regs->tmroff_h, hi);
+}
+
/* Caller must hold ptp_qoriq->lock. */
static void set_alarm(struct ptp_qoriq *ptp_qoriq)
{
@@ -55,7 +78,9 @@ static void set_alarm(struct ptp_qoriq *ptp_qoriq)
u64 ns;
u32 lo, hi;
- ns = tmr_cnt_read(ptp_qoriq) + 1500000000ULL;
+ ns = tmr_cnt_read(ptp_qoriq) + tmr_offset_read(ptp_qoriq)
+ + 1500000000ULL;
+
ns = div_u64(ns, 1000000000UL) * 1000000000ULL;
ns -= ptp_qoriq->tclk_period;
hi = ns >> 32;
@@ -207,15 +232,24 @@ EXPORT_SYMBOL_GPL(ptp_qoriq_adjfine);
int ptp_qoriq_adjtime(struct ptp_clock_info *ptp, s64 delta)
{
- s64 now;
- unsigned long flags;
struct ptp_qoriq *ptp_qoriq = container_of(ptp, struct ptp_qoriq, caps);
+ s64 now, curr_delta;
+ unsigned long flags;
spin_lock_irqsave(&ptp_qoriq->lock, flags);
- now = tmr_cnt_read(ptp_qoriq);
- now += delta;
- tmr_cnt_write(ptp_qoriq, now);
+ /* On LS1021A, eTSEC2 and eTSEC3 do not take into account the TMR_OFF
+ * adjustment
+ */
+ if (ptp_qoriq->etsec) {
+ now = tmr_cnt_read(ptp_qoriq);
+ now += delta;
+ tmr_cnt_write(ptp_qoriq, now);
+ } else {
+ curr_delta = tmr_offset_read(ptp_qoriq);
+ curr_delta += delta;
+ tmr_offset_write(ptp_qoriq, curr_delta);
+ }
set_fipers(ptp_qoriq);
spin_unlock_irqrestore(&ptp_qoriq->lock, flags);
@@ -232,7 +266,7 @@ int ptp_qoriq_gettime(struct ptp_clock_info *ptp, struct timespec64 *ts)
spin_lock_irqsave(&ptp_qoriq->lock, flags);
- ns = tmr_cnt_read(ptp_qoriq);
+ ns = tmr_cnt_read(ptp_qoriq) + tmr_offset_read(ptp_qoriq);
spin_unlock_irqrestore(&ptp_qoriq->lock, flags);
@@ -253,6 +287,7 @@ int ptp_qoriq_settime(struct ptp_clock_info *ptp,
spin_lock_irqsave(&ptp_qoriq->lock, flags);
+ tmr_offset_write(ptp_qoriq, 0);
tmr_cnt_write(ptp_qoriq, ns);
set_fipers(ptp_qoriq);
@@ -488,6 +523,7 @@ int ptp_qoriq_init(struct ptp_qoriq *ptp_qoriq, void __iomem *base,
/* The eTSEC uses differnt memory map with DPAA/ENETC */
if (of_device_is_compatible(node, "fsl,etsec-ptp")) {
+ ptp_qoriq->etsec = true;
ptp_qoriq->regs.ctrl_regs = base + ETSEC_CTRL_REGS_OFFSET;
ptp_qoriq->regs.alarm_regs = base + ETSEC_ALARM_REGS_OFFSET;
ptp_qoriq->regs.fiper_regs = base + ETSEC_FIPER_REGS_OFFSET;
diff --git a/include/linux/fsl/ptp_qoriq.h b/include/linux/fsl/ptp_qoriq.h
index 01acebe37fab..b301bf7199d3 100644
--- a/include/linux/fsl/ptp_qoriq.h
+++ b/include/linux/fsl/ptp_qoriq.h
@@ -149,6 +149,7 @@ struct ptp_qoriq {
struct device *dev;
bool extts_fifo_support;
bool fiper3_support;
+ bool etsec;
int irq;
int phc_index;
u32 tclk_period; /* nanoseconds */
--
2.17.1
On Fri, Jan 20, 2023 at 02:10:34AM +0530, [email protected] wrote:
> From: Nikhil Gupta <[email protected]>
>
> 1588 driver loses about 1us in adjtime operation at PTP slave
> This is because adjtime operation uses a slow non-atomic tmr_cnt_read()
> followed by tmr_cnt_write() operation.
>
> In the above sequence, since the timer counter operation keeps
> incrementing, it leads to latency. The tmr_offset register
> (which is added to TMR_CNT_H/L register giving the current time)
> must be programmed with the delta nanoseconds.
>
> Signed-off-by: Nikhil Gupta <[email protected]>
> ---
Reviewed-by: Vladimir Oltean <[email protected]>
Tested-by: Vladimir Oltean <[email protected]>
One detail regarding the process: you should have clarified your
expectations, i.e. the patch should have explicitly been generated with
git send-email --subject-prefix='PATCH v2 net-next' since it is not
targeting the 'net.git' tree for bug fixes on the master branch (which
are also backported to stable trees), but rather will be included in the
pull request for the upcoming 6.3 kernel.
Hello:
This patch was applied to netdev/net-next.git (master)
by Jakub Kicinski <[email protected]>:
On Fri, 20 Jan 2023 02:10:34 +0530 you wrote:
> From: Nikhil Gupta <[email protected]>
>
> 1588 driver loses about 1us in adjtime operation at PTP slave
> This is because adjtime operation uses a slow non-atomic tmr_cnt_read()
> followed by tmr_cnt_write() operation.
>
> In the above sequence, since the timer counter operation keeps
> incrementing, it leads to latency. The tmr_offset register
> (which is added to TMR_CNT_H/L register giving the current time)
> must be programmed with the delta nanoseconds.
>
> [...]
Here is the summary with links:
- [v2] ptp_qoriq: fix latency in ptp_qoriq_adjtime() operation
https://git.kernel.org/netdev/net-next/c/24a7fffb2533
You are awesome, thank you!
--
Deet-doot-dot, I am a bot.
https://korg.docs.kernel.org/patchwork/pwbot.html