RE: [PATCH v4 2/2] misc: xlnx_vcu: Add Xilinx ZYNQMP VCU logicoreIP init driver

From: Dhaval Rajeshbhai Shah
Date: Mon Dec 18 2017 - 00:38:27 EST


Hi Randy,

> -----Original Message-----
> From: Randy Dunlap [mailto:rdunlap@xxxxxxxxxxxxx]
> Sent: Sunday, December 17, 2017 9:40 AM
> To: Dhaval Rajeshbhai Shah <DSHAH@xxxxxxxxxx>; arnd@xxxxxxxx;
> gregkh@xxxxxxxxxxxxxxxxxxx; robh+dt@xxxxxxxxxx; mark.rutland@xxxxxxx
> Cc: devicetree@xxxxxxxxxxxxxxx; linux-kernel@xxxxxxxxxxxxxxx;
> michal.simek@xxxxxxxxxx; Hyun Kwon <hyunk@xxxxxxxxxx>
> Subject: Re: [PATCH v4 2/2] misc: xlnx_vcu: Add Xilinx ZYNQMP VCU logicoreIP
> init driver
>
> On 12/16/2017 10:07 PM, Dhaval Rajeshbhai Shah wrote:
> > Hi Randy,
> >
> > Thanks a lot for the review.
> >
> >> -----Original Message-----
> >> From: Randy Dunlap [mailto:rdunlap@xxxxxxxxxxxxx]
> >> Sent: Saturday, December 16, 2017 2:18 PM
> >> To: Dhaval Rajeshbhai Shah <DSHAH@xxxxxxxxxx>; arnd@xxxxxxxx;
> >> gregkh@xxxxxxxxxxxxxxxxxxx; robh+dt@xxxxxxxxxx; mark.rutland@xxxxxxx
> >> Cc: devicetree@xxxxxxxxxxxxxxx; linux-kernel@xxxxxxxxxxxxxxx;
> >> michal.simek@xxxxxxxxxx; Hyun Kwon <hyunk@xxxxxxxxxx>; Dhaval
> >> Rajeshbhai Shah <DSHAH@xxxxxxxxxx>
> >> Subject: Re: [PATCH v4 2/2] misc: xlnx_vcu: Add Xilinx ZYNQMP VCU
> >> logicoreIP init driver
> >>
> >> On 12/14/2017 11:24 PM, Dhaval Shah wrote:
> >>> Xilinx ZYNQMP logicoreIP Init driver is based on the new LogiCoreIP
> >>> design created. This driver provides the processing system and
> >>> programmable logic isolation. Set the frequency based on the clock
> >>> information get from the logicoreIP register set.
> >>>
> >>> It is put in drivers/misc as there is no subsystem for this logicoreIP.
> >>>
> >>> Signed-off-by: Dhaval Shah <dshah@xxxxxxxxxx>
> >>> ---
> >>>
> >>> drivers/misc/Kconfig | 15 ++
> >>> drivers/misc/Makefile | 1 +
> >>> drivers/misc/xlnx_vcu.c | 631
> >>> ++++++++++++++++++++++++++++++++++++++++++++++++
> >>> 3 files changed, 647 insertions(+)
> >>> create mode 100644 drivers/misc/xlnx_vcu.c
> >>
> >>> diff --git a/drivers/misc/xlnx_vcu.c b/drivers/misc/xlnx_vcu.c new
> >>> file mode 100644 index 0000000..f489d34
> >>> --- /dev/null
> >>> +++ b/drivers/misc/xlnx_vcu.c
> >>> @@ -0,0 +1,631 @@
> >>> +// SPDX-License-Identifier: GPL-2.0
> >>> +/*
> >>> + * Xilinx VCU Init
> >>> + *
> >>> + * Copyright (C) 2016 - 2017 Xilinx, Inc.
> >>> + *
> >>> + * Contacts Dhaval Shah <dshah@xxxxxxxxxx>
> >>> + */
> >>> +#include <linux/clk.h>
> >>> +#include <linux/device.h>
> >>> +#include <linux/errno.h>
> >>> +#include <linux/io.h>
> >>> +#include <linux/module.h>
> >>> +#include <linux/of_platform.h>
> >>> +#include <linux/platform_device.h>
> >>
> >> [snip]
> >>
> >>
> >>> +/**
> >>> + * xvcu_set_vcu_pll_info - Set the VCU PLL info
> >>> + * @xvcu: Pointer to the xvcu_device structure
> >>> + *
> >>> + * Programming the VCU PLL based on the user configuration
> >>> + * (ref clock freq, core clock freq, mcu clock freq).
> >>> + * Core clock frequency has higher priority than mcu clock
> >>> +frequency
> >>> + * Errors in following cases
> >>> + * - When mcu or clock clock get from logicoreIP is 0
> >>> + * - When VCU PLL DIV related bits value other than 1
> >>> + * - When proper data not found for given data
> >>> + * - When sis570_1 clocksource related operation failed
> >>> + *
> >>> + * Return: Returns status, either success or error+reason
> >>> + */
> >>> +static int xvcu_set_vcu_pll_info(struct xvcu_device *xvcu) {
> >>> + u32 refclk, coreclk, mcuclk, inte, deci;
> >>> + u32 divisor_mcu, divisor_core, fvco;
> >>> + u32 clkoutdiv, vcu_pll_ctrl, pll_clk;
> >>> + u32 cfg_val, mod, ctrl;
> >>> + int ret;
> >>> + unsigned int i;
> >>> + const struct xvcu_pll_cfg *found = NULL;
> >>> +
> >>> + inte = xvcu_read(xvcu->logicore_reg_ba, VCU_PLL_CLK);
> >>> + deci = xvcu_read(xvcu->logicore_reg_ba, VCU_PLL_CLK_DEC);
> >>> + coreclk = xvcu_read(xvcu->logicore_reg_ba, VCU_CORE_CLK) * MHZ;
> >>> + mcuclk = xvcu_read(xvcu->logicore_reg_ba, VCU_MCU_CLK) * MHZ;
> >>> + if (!mcuclk || !coreclk) {
> >>> + dev_err(xvcu->dev, "Invalid mcu and core clock data\n");
> >>> + return -EINVAL;
> >>> + }
> >>> +
> >>> + refclk = (inte * MHZ) + (deci * (MHZ / FRAC));
> >>> + dev_dbg(xvcu->dev, "Ref clock from logicoreIP is %uHz\n", refclk);
> >>> + dev_dbg(xvcu->dev, "Core clock from logicoreIP is %uHz\n", coreclk);
> >>> + dev_dbg(xvcu->dev, "Mcu clock from logicoreIP is %uHz\n", mcuclk);
> >>> +
> >>> + clk_disable_unprepare(xvcu->pll_ref);
> >>> + ret = clk_set_rate(xvcu->pll_ref, refclk);
> >>> + if (ret)
> >>> + dev_warn(xvcu->dev, "failed to set logicoreIP refclk rate\n");
> >>> +
> >>> + ret = clk_prepare_enable(xvcu->pll_ref);
> >>> + if (ret) {
> >>> + dev_err(xvcu->dev, "failed to enable pll_ref clock source\n");
> >>> + return ret;
> >>> + }
> >>> +
> >>> + refclk = clk_get_rate(xvcu->pll_ref);
> >>> +
> >>> + /*
> >>> + * The divide-by-2 should be always enabled (==1)
> >>> + * to meet the timing in the design.
> >>> + * Otherwise, it's an error
> >>> + */
> >>> + vcu_pll_ctrl = xvcu_read(xvcu->vcu_slcr_ba, VCU_PLL_CTRL);
> >>> + clkoutdiv = vcu_pll_ctrl >> VCU_PLL_CTRL_CLKOUTDIV_SHIFT;
> >>> + clkoutdiv = clkoutdiv && VCU_PLL_CTRL_CLKOUTDIV_MASK;
> >>> + if (clkoutdiv != 1) {
> >>> + dev_err(xvcu->dev, "clkoutdiv value is invalid\n");
> >>> + return -EINVAL;
> >>> + }
> >>> +
> >>> + for (i = ARRAY_SIZE(xvcu_pll_cfg) - 1; i > 0; i--) {
> >>
> >> When does that for loop terminate?
> >>
> > This loop will terminate either it reach to i =0 and one other case is when it
> find the proper expected value. Break statement is used to exit in that case.
>
> Hm, well, I did mis-read it, but I think that there is still an off-by-one problem.
>
> Consider the case of the value is not found in the array.
>
> If ARRAY_SIZE(xvcu_pll_cfg) is 5:
>
> for (i = ARRAY_SIZE(xvcu_pll_cfg) - 1; i > 0; i--) {
>
> i == 4:
> check xvcu_pll_cfg[4]
> i--;
>
> i == 3:
> check xvcu_pll_cfg[3]
> i--;
>
> i == 2;
> check xvcu_pll_cfg[2]
> i--;
>
> i == 1:
> check xvcu_pll_cfg[1]
> i--;
>
> i == 0;
> terminate for-loop;
> xvcu_pll_cfg[0] is never checked;
>
> Is this clear?
>
Yes. You are right. It will miss index 0 in that case. I will update and send the next patch set.
>
> >>> + const struct xvcu_pll_cfg *cfg = &xvcu_pll_cfg[i];
> >>> +
> >>> + fvco = cfg->fbdiv * refclk;
> >>> + if (fvco >= FVCO_MIN && fvco <= FVCO_MAX) {
> >>> + pll_clk = fvco / VCU_PLL_DIV2;
> >>> + if (fvco % VCU_PLL_DIV2 != 0)
> >>> + pll_clk++;
> >>> + mod = pll_clk % coreclk;
> >>> + if (mod < LIMIT) {
> >>> + divisor_core = pll_clk / coreclk;
> >>> + } else if (coreclk - mod < LIMIT) {
> >>> + divisor_core = pll_clk / coreclk;
> >>> + divisor_core++;
> >>> + } else {
> >>> + continue;
> >>> + }
> >>> + if (divisor_core >= DIVISOR_MIN &&
> >>> + divisor_core <= DIVISOR_MAX) {
> >>> + found = cfg;
> >>> + divisor_mcu = pll_clk / mcuclk;
> >>> + mod = pll_clk % mcuclk;
> >>> + if (mcuclk - mod < LIMIT)
> >>> + divisor_mcu++;
> >>> + break;
> > This is the break statement to exit when proper value is find.
> >>> + }
> >>> + }
> >>> + }
> >>> +
> >>> + if (!found) {
> >>> + dev_err(xvcu->dev, "Invalid clock combination.\n");
> >>> + return -EINVAL;
> >>> + }
> >>
> >> [snip]
>
>
> --
> ~Randy