| Message ID | 20220228203957.367371-1-jonathan.lemon@gmail.com (mailing list archive) |
|---|---|
| State | Accepted |
| Commit | 90f8f4c0e3cebd541deaa45cf0e470bb9810dd4f |
| Delegated to: | Netdev Maintainers |
| Headers | show |
| Series | [net-next] ptp: ocp: Add ptp_ocp_adjtime_coarse for large adjustments | expand |
On Mon, Feb 28, 2022 at 12:39:57PM -0800, Jonathan Lemon wrote: > In ("ptp: ocp: Have FPGA fold in ns adjustment for adjtime."), the > ns adjustment was written to the FPGA register, so the clock could > accurately perform adjustments. > > However, the adjtime() call passes in a s64, while the clock adjustment > registers use a s32. When trying to perform adjustments with a large > value (37 sec), things fail. > > Examine the incoming delta, and if larger than 1 sec, use the original > (coarse) adjustment method. If smaller than 1 sec, then allow the > FPGA to fold in the changes over a 1 second window. > > Fixes: 6d59d4fa1789 ("ptp: ocp: Have FPGA fold in ns adjustment for adjtime.") > Signed-off-by: Jonathan Lemon <jonathan.lemon@gmail.com> Acked-by: Richard Cochran <richardcochran@gmail.com>
On Mon, 28 Feb 2022 12:39:57 -0800 Jonathan Lemon wrote: > In ("ptp: ocp: Have FPGA fold in ns adjustment for adjtime."), the > ns adjustment was written to the FPGA register, so the clock could > accurately perform adjustments. > > However, the adjtime() call passes in a s64, while the clock adjustment > registers use a s32. When trying to perform adjustments with a large > value (37 sec), things fail. > > Examine the incoming delta, and if larger than 1 sec, use the original > (coarse) adjustment method. If smaller than 1 sec, then allow the > FPGA to fold in the changes over a 1 second window. > > Fixes: 6d59d4fa1789 ("ptp: ocp: Have FPGA fold in ns adjustment for adjtime.") > Signed-off-by: Jonathan Lemon <jonathan.lemon@gmail.com> This one's tagged for net-next - do you intend for it to go to net-next, or is that a typo?
On Tue, Mar 01, 2022 at 06:21:53PM -0800, Jakub Kicinski wrote: > On Mon, 28 Feb 2022 12:39:57 -0800 Jonathan Lemon wrote: > > In ("ptp: ocp: Have FPGA fold in ns adjustment for adjtime."), the > > ns adjustment was written to the FPGA register, so the clock could > > accurately perform adjustments. > > > > However, the adjtime() call passes in a s64, while the clock adjustment > > registers use a s32. When trying to perform adjustments with a large > > value (37 sec), things fail. > > > > Examine the incoming delta, and if larger than 1 sec, use the original > > (coarse) adjustment method. If smaller than 1 sec, then allow the > > FPGA to fold in the changes over a 1 second window. > > > > Fixes: 6d59d4fa1789 ("ptp: ocp: Have FPGA fold in ns adjustment for adjtime.") > > Signed-off-by: Jonathan Lemon <jonathan.lemon@gmail.com> > > This one's tagged for net-next - do you intend for it to go to net-next, > or is that a typo? I build and test net-next, so that was my target.
On Wed, 2 Mar 2022 08:48:00 -0800 Jonathan Lemon wrote: > > This one's tagged for net-next - do you intend for it to go to net-next, > > or is that a typo? > > I build and test net-next, so that was my target. You should develop fixes against the current tree, not -next trees.
Hello: This patch was applied to netdev/net.git (master) by Jakub Kicinski <kuba@kernel.org>: On Mon, 28 Feb 2022 12:39:57 -0800 you wrote: > In ("ptp: ocp: Have FPGA fold in ns adjustment for adjtime."), the > ns adjustment was written to the FPGA register, so the clock could > accurately perform adjustments. > > However, the adjtime() call passes in a s64, while the clock adjustment > registers use a s32. When trying to perform adjustments with a large > value (37 sec), things fail. > > [...] Here is the summary with links: - [net-next] ptp: ocp: Add ptp_ocp_adjtime_coarse for large adjustments https://git.kernel.org/netdev/net/c/90f8f4c0e3ce You are awesome, thank you!
diff --git a/drivers/ptp/ptp_ocp.c b/drivers/ptp/ptp_ocp.c index 0f1b5a7d2a89..17ad5f0d13b2 100644 --- a/drivers/ptp/ptp_ocp.c +++ b/drivers/ptp/ptp_ocp.c @@ -607,7 +607,7 @@ ptp_ocp_settime(struct ptp_clock_info *ptp_info, const struct timespec64 *ts) } static void -__ptp_ocp_adjtime_locked(struct ptp_ocp *bp, u64 adj_val) +__ptp_ocp_adjtime_locked(struct ptp_ocp *bp, u32 adj_val) { u32 select, ctrl; @@ -615,7 +615,7 @@ __ptp_ocp_adjtime_locked(struct ptp_ocp *bp, u64 adj_val) iowrite32(OCP_SELECT_CLK_REG, &bp->reg->select); iowrite32(adj_val, &bp->reg->offset_ns); - iowrite32(adj_val & 0x7f, &bp->reg->offset_window_ns); + iowrite32(NSEC_PER_SEC, &bp->reg->offset_window_ns); ctrl = OCP_CTRL_ADJUST_OFFSET | OCP_CTRL_ENABLE; iowrite32(ctrl, &bp->reg->ctrl); @@ -624,6 +624,22 @@ __ptp_ocp_adjtime_locked(struct ptp_ocp *bp, u64 adj_val) iowrite32(select >> 16, &bp->reg->select); } +static void +ptp_ocp_adjtime_coarse(struct ptp_ocp *bp, u64 delta_ns) +{ + struct timespec64 ts; + unsigned long flags; + int err; + + spin_lock_irqsave(&bp->lock, flags); + err = __ptp_ocp_gettime_locked(bp, &ts, NULL); + if (likely(!err)) { + timespec64_add_ns(&ts, delta_ns); + __ptp_ocp_settime_locked(bp, &ts); + } + spin_unlock_irqrestore(&bp->lock, flags); +} + static int ptp_ocp_adjtime(struct ptp_clock_info *ptp_info, s64 delta_ns) { @@ -631,6 +647,11 @@ ptp_ocp_adjtime(struct ptp_clock_info *ptp_info, s64 delta_ns) unsigned long flags; u32 adj_ns, sign; + if (delta_ns > NSEC_PER_SEC || -delta_ns > NSEC_PER_SEC) { + ptp_ocp_adjtime_coarse(bp, delta_ns); + return 0; + } + sign = delta_ns < 0 ? BIT(31) : 0; adj_ns = sign ? -delta_ns : delta_ns;
In ("ptp: ocp: Have FPGA fold in ns adjustment for adjtime."), the ns adjustment was written to the FPGA register, so the clock could accurately perform adjustments. However, the adjtime() call passes in a s64, while the clock adjustment registers use a s32. When trying to perform adjustments with a large value (37 sec), things fail. Examine the incoming delta, and if larger than 1 sec, use the original (coarse) adjustment method. If smaller than 1 sec, then allow the FPGA to fold in the changes over a 1 second window. Fixes: 6d59d4fa1789 ("ptp: ocp: Have FPGA fold in ns adjustment for adjtime.") Signed-off-by: Jonathan Lemon <jonathan.lemon@gmail.com> --- drivers/ptp/ptp_ocp.c | 25 +++++++++++++++++++++++-- 1 file changed, 23 insertions(+), 2 deletions(-)