| lists.openwall.net | lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC | |
|
Open Source and information security mailing list archives
| ||
|
Message-ID: <20160630154302.GB4279@ulmo.ba.sec>
Date: Thu, 30 Jun 2016 17:43:02 +0200
From: Thierry Reding <thierry.reding@...il.com>
To: Rhyland Klein <rklein@...dia.com>
Cc: Jon Hunter <jonathanh@...dia.com>,
Peter De Schrijver <pdeschrijver@...dia.com>,
Mike Turquette <mturquette@...libre.com>,
Stephen Warren <swarren@...dotorg.org>,
Stephen Boyd <sboyd@...eaurora.org>,
Alexandre Courbot <gnurou@...il.com>,
linux-clk@...r.kernel.org, linux-tegra@...r.kernel.org,
linux-kernel@...r.kernel.org,
Andrew Bresticker <abrestic@...omium.org>
Subject: Re: [PATCH v5] clk: tegra: Initialize UTMIPLL when enabling PLLU
On Thu, Jun 30, 2016 at 11:40:19AM -0400, Rhyland Klein wrote:
> On 6/30/2016 11:37 AM, Thierry Reding wrote:
> > * PGP Signed by an unknown key
> >
> > On Thu, Jun 30, 2016 at 11:32:14AM -0400, Rhyland Klein wrote:
> >> On 6/17/2016 11:23 AM, Thierry Reding wrote:
> >>>> Old Signed by an unknown key
> >>>
> >>> On Fri, Jun 17, 2016 at 02:49:41PM +0100, Jon Hunter wrote:
> >>>> Hi Thierry,
> >>>>
> >>>> On 26/05/16 17:41, Rhyland Klein wrote:
> >>>>> From: Andrew Bresticker <abrestic@...omium.org>
> >>>>>
> >>>>> Move the UTMIPLL initialization code form clk-tegra<chip>.c files into
> >>>>> clk-pll.c. UTMIPLL was being configured and set in HW control right
> >>>>> after registration. However, when the clock init_table is processed and
> >>>>> child clks of PLLU are enabled, it will call in and enable PLLU as
> >>>>> well, and initiate SW enabling sequence even though PLLU is already in
> >>>>> HW control. This leads to getting UTMIPLL stuck with a SEQ_BUSY status.
> >>>>>
> >>>>> Doing the initialization once during pllu_enable means we configure it
> >>>>> properly into HW control.
> >>>>>
> >>>>> A side effect of the commonization/localization of the UTMIPLL init
> >>>>> code, is that it corrects some errors that were present for earlier
> >>>>> generations. For instance, in clk-tegra124.c, it used to have:
> >>>>>
> >>>>> define UTMIP_PLL_CFG1_ENABLE_DLY_COUNT(x) (((x) & 0x1f) << 6)
> >>>>>
> >>>>> when the correct shift to use is present in the new version:
> >>>>>
> >>>>> define UTMIP_PLL_CFG1_ENABLE_DLY_COUNT(x) (((x) & 0x1f) << 27)
> >>>>>
> >>>>> which matches the Tegra124 TRM register definition.
> >>>>>
> >>>>> Signed-off-by: Andrew Bresticker <abrestic@...omium.org>
> >>>>>
> >>>>> [rklein: Merged in some later fixes for potential deadlocks]
> >>>>>
> >>>>> Signed-off-by: Rhyland Klein <rklein@...dia.com>
> >>>>> ---
> >>>>> v5:
> >>>>> - Initialized flags to 0 to avoid harmless spinlock warnings
> >>>>>
> >>>>> v4:
> >>>>> - Re-added examples in patch description
> >>>>>
> >>>>> v3:
> >>>>> - Flushed out description to describe this patch.
> >>>>>
> >>>>> drivers/clk/tegra/clk-pll.c | 484 +++++++++++++++++++++++++++++++++++++++
> >>>>> drivers/clk/tegra/clk-tegra114.c | 155 +------------
> >>>>> drivers/clk/tegra/clk-tegra124.c | 156 +------------
> >>>>> drivers/clk/tegra/clk-tegra210.c | 182 +--------------
> >>>>> drivers/clk/tegra/clk-tegra30.c | 113 +--------
> >>>>> drivers/clk/tegra/clk.h | 17 ++
> >>>>> 6 files changed, 510 insertions(+), 597 deletions(-)
> >>>>>
> >>>>> diff --git a/drivers/clk/tegra/clk-pll.c b/drivers/clk/tegra/clk-pll.c
> >>>>> index 4e194ecc8d5e..31e20110fae4 100644
> >>>>> --- a/drivers/clk/tegra/clk-pll.c
> >>>>> +++ b/drivers/clk/tegra/clk-pll.c
> >>>>
> >>>> ...
> >>>>
> >>>>> +static int clk_pllu_tegra210_enable(struct clk_hw *hw)
> >>>>> +{
> >>>>> + struct tegra_clk_pll *pll = to_clk_pll(hw);
> >>>>> + struct clk_hw *pll_ref = clk_hw_get_parent(hw);
> >>>>> + struct clk_hw *osc = clk_hw_get_parent(pll_ref);
> >>>>> + unsigned long flags = 0, input_rate;
> >>>>> + unsigned int i;
> >>>>> + int ret = 0;
> >>>>> + u32 val;
> >>>>> +
> >>>>> + if (!osc) {
> >>>>> + pr_err("%s: failed to get OSC clock\n", __func__);
> >>>>> + return -EINVAL;
> >>>>> + }
> >>>>> + input_rate = clk_hw_get_rate(osc);
> >>>>> +
> >>>>> + if (pll->lock)
> >>>>> + spin_lock_irqsave(pll->lock, flags);
> >>>>> +
> >>>>> + _clk_pll_enable(hw);
> >>>>> + ret = clk_pll_wait_for_lock(pll);
> >>>>> + if (ret < 0)
> >>>>> + goto out;
> >>>>> +
> >>>>> + for (i = 0; i < ARRAY_SIZE(utmi_parameters); i++) {
> >>>>> + if (input_rate == utmi_parameters[i].osc_frequency)
> >>>>> + break;
> >>>>> + }
> >>>>> +
> >>>>> + if (i == ARRAY_SIZE(utmi_parameters)) {
> >>>>> + pr_err("%s: Unexpected input rate %lu\n", __func__, input_rate);
> >>>>> + ret = -EINVAL;
> >>>>> + goto out;
> >>>>> + }
> >>>>> +
> >>>>> + val = pll_readl_base(pll);
> >>>>> + val &= ~PLLU_BASE_OVERRIDE;
> >>>>> + pll_writel_base(val, pll);
> >>>>> +
> >>>>> + /* Put PLLU under HW control */
> >>>>> + val = readl_relaxed(pll->clk_base + PLLU_HW_PWRDN_CFG0);
> >>>>> + val |= PLLU_HW_PWRDN_CFG0_IDDQ_PD_INCLUDE |
> >>>>> + PLLU_HW_PWRDN_CFG0_USE_SWITCH_DETECT |
> >>>>> + PLLU_HW_PWRDN_CFG0_USE_LOCKDET;
> >>>>> + val &= ~(PLLU_HW_PWRDN_CFG0_CLK_ENABLE_SWCTL |
> >>>>> + PLLU_HW_PWRDN_CFG0_CLK_SWITCH_SWCTL);
> >>>>> + writel_relaxed(val, pll->clk_base + PLLU_HW_PWRDN_CFG0);
> >>>>> +
> >>>>> + val = readl_relaxed(pll->clk_base + XUSB_PLL_CFG0);
> >>>>> + val &= ~XUSB_PLL_CFG0_PLLU_LOCK_DLY;
> >>>>> + writel_relaxed(val, pll->clk_base + XUSB_PLL_CFG0);
> >>>>> + udelay(1);
> >>>>> +
> >>>>> + val = readl_relaxed(pll->clk_base + PLLU_HW_PWRDN_CFG0);
> >>>>> + val |= PLLU_HW_PWRDN_CFG0_SEQ_ENABLE;
> >>>>> + writel_relaxed(val, pll->clk_base + PLLU_HW_PWRDN_CFG0);
> >>>>> + udelay(1);
> >>>>> +
> >>>>> + /* Disable PLLU clock branch to UTMIPLL since it uses OSC */
> >>>>> + val = pll_readl_base(pll);
> >>>>> + val &= ~PLLU_BASE_CLKENABLE_USB;
> >>>>> + pll_writel_base(val, pll);
> >>>>> +
> >>>>> + val = readl_relaxed(pll->clk_base + UTMIPLL_HW_PWRDN_CFG0);
> >>>>> + if (val & UTMIPLL_HW_PWRDN_CFG0_SEQ_ENABLE) {
> >>>>> + pr_debug("UTMIPLL already enabled\n");
> >>>>> + goto out;
> >>>>> + }
> >>>>> + val &= ~UTMIPLL_HW_PWRDN_CFG0_IDDQ_OVERRIDE;
> >>>>> + writel_relaxed(val, pll->clk_base + UTMIPLL_HW_PWRDN_CFG0);
> >>>>> +
> >>>>> + /* Program UTMIP PLL stable and active counts */
> >>>>> + val = readl_relaxed(pll->clk_base + UTMIP_PLL_CFG2);
> >>>>> + val &= ~UTMIP_PLL_CFG2_STABLE_COUNT(~0);
> >>>>> + val |= UTMIP_PLL_CFG2_STABLE_COUNT(utmi_parameters[i].stable_count);
> >>>>> + val &= ~UTMIP_PLL_CFG2_ACTIVE_DLY_COUNT(~0);
> >>>>> + val |= UTMIP_PLL_CFG2_ACTIVE_DLY_COUNT(
> >>>>> + utmi_parameters[i].active_delay_count);
> >>>>> + val |= UTMIP_PLL_CFG2_PHY_XTAL_CLOCKEN;
> >>>>> + writel_relaxed(val, pll->clk_base + UTMIP_PLL_CFG2);
> >>>>> +
> >>>>> + /* Program UTMIP PLL delay and oscillator frequency counts */
> >>>>> + val = readl_relaxed(pll->clk_base + UTMIP_PLL_CFG1);
> >>>>> + val &= ~UTMIP_PLL_CFG1_ENABLE_DLY_COUNT(~0);
> >>>>> + val |= UTMIP_PLL_CFG1_ENABLE_DLY_COUNT(
> >>>>> + utmi_parameters[i].enable_delay_count);
> >>>>> + val &= ~UTMIP_PLL_CFG1_XTAL_FREQ_COUNT(~0);
> >>>>> + val |= UTMIP_PLL_CFG1_XTAL_FREQ_COUNT(
> >>>>> + utmi_parameters[i].xtal_freq_count);
> >>>>> + writel_relaxed(val, pll->clk_base + UTMIP_PLL_CFG1);
> >>>>> +
> >>>>> + /* Remove power downs from UTMIP PLL control bits */
> >>>>> + val = readl_relaxed(pll->clk_base + UTMIP_PLL_CFG1);
> >>>>> + val &= ~UTMIP_PLL_CFG1_FORCE_PLL_ENABLE_POWERDOWN;
> >>>>> + val |= UTMIP_PLL_CFG1_FORCE_PLL_ENABLE_POWERUP;
> >>>>> + writel_relaxed(val, pll->clk_base + UTMIP_PLL_CFG1);
> >>>>> + udelay(100);
> >>>>
> >>>> In next-20160617 I see that this udelay is now a usleep_range(100, 200)
> >>>> and this is causing the following splat when the clock is enabled. I
> >>>> don't think that we can use usleep here ...
> >>>
> >>> Okay, I'll back out the patch. I'd really prefer to avoid busy-looping
> >>> for 100 microseconds here, so can we please find another way to do this?
> >>>
> >>
> >> It looks like we should be able to use a short udelay of 1-2us. I
> >> believe the original code had udelay(1) and I know Jon and I tested
> >> udelay(2) and it was ok.
> >
> > What original code? The downstream driver? If so I'd be leaning towards
> > simply adopting that. Everything else in this functions seems to want to
> > wait for 1 us, seems natural for this to do as well.
>
> Sorry I wasn't clear. The code in the clk-tegraXX specific drivers was
> using udelay(1) as you pointed out, thats what I meant.
On further looking at the downstream code, the write before the udelay()
above should probably also use the non-relaxed version because the code
in our downstream kernel uses a dsb() after the write (which makes it
equivalent to a plain writel()).
Thierry
Download attachment "signature.asc" of type "application/pgp-signature" (820 bytes)
Powered by blists - more mailing lists