| 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: <f5bc9590-f37e-491e-9978-c1eab8914c30@salutedevices.com>
Date: Thu, 18 Jul 2024 14:01:28 +0300
From: Jan Dakinevich <jan.dakinevich@...utedevices.com>
To: Jerome Brunet <jbrunet@...libre.com>
CC: Philipp Zabel <p.zabel@...gutronix.de>, Stephen Boyd <sboyd@...nel.org>,
Neil Armstrong <neil.armstrong@...aro.org>, <linux-kernel@...r.kernel.org>,
<linux-amlogic@...ts.infradead.org>, <linux-clk@...r.kernel.org>
Subject: Re: [PATCH 1/8] reset: amlogic: convert driver to regmap
On 7/18/24 10:19, Jerome Brunet wrote:
> On Thu 18 Jul 2024 at 05:39, Jan Dakinevich <jan.dakinevich@...utedevices.com> wrote:
>
>> On 7/10/24 19:25, Jerome Brunet wrote:
>>> To allow using the same driver for the main reset controller and the
>>> auxiliary ones embedded in the clock controllers, convert the
>>> the Amlogic reset driver to regmap.
>>>
>>> Signed-off-by: Jerome Brunet <jbrunet@...libre.com>
>>> ---
>>> drivers/reset/reset-meson.c | 80 ++++++++++++++++++++-----------------
>>> 1 file changed, 44 insertions(+), 36 deletions(-)
>>>
>>> diff --git a/drivers/reset/reset-meson.c b/drivers/reset/reset-meson.c
>>> index f78be97898bc..8f3d6e9df235 100644
>>> --- a/drivers/reset/reset-meson.c
>>> +++ b/drivers/reset/reset-meson.c
>>> @@ -11,36 +11,44 @@
>>> #include <linux/of.h>
>>> #include <linux/module.h>
>>> #include <linux/platform_device.h>
>>> +#include <linux/regmap.h>
>>> #include <linux/reset-controller.h>
>>> #include <linux/slab.h>
>>> #include <linux/types.h>
>>>
>>> -#define BITS_PER_REG 32
>>> -
>>> struct meson_reset_param {
>>> int reg_count;
>>> int level_offset;
>>> };
>>>
>>> struct meson_reset {
>>> - void __iomem *reg_base;
>>> const struct meson_reset_param *param;
>>> struct reset_controller_dev rcdev;
>>> - spinlock_t lock;
>>> + struct regmap *map;
>>> };
>>>
>>> +static void meson_reset_offset_and_bit(struct meson_reset *data,
>>> + unsigned long id,
>>> + unsigned int *offset,
>>> + unsigned int *bit)
>>> +{
>>> + unsigned int stride = regmap_get_reg_stride(data->map);
>>> +
>>> + *offset = (id / (stride * BITS_PER_BYTE)) * stride;
>>> + *bit = id % (stride * BITS_PER_BYTE);
>>> +}
>>> +
>>> static int meson_reset_reset(struct reset_controller_dev *rcdev,
>>> - unsigned long id)
>>> + unsigned long id)
>>> {
>>> struct meson_reset *data =
>>> container_of(rcdev, struct meson_reset, rcdev);
>>> - unsigned int bank = id / BITS_PER_REG;
>>> - unsigned int offset = id % BITS_PER_REG;
>>> - void __iomem *reg_addr = data->reg_base + (bank << 2);
>>> + unsigned int offset, bit;
>>>
>>> - writel(BIT(offset), reg_addr);
>>> + meson_reset_offset_and_bit(data, id, &offset, &bit);
>>>
>>> - return 0;
>>> + return regmap_update_bits(data->map, offset,
>>> + BIT(bit), BIT(bit));
>>
>> regmap_update_bits() is not equal to writel(), I suppose regmap_write()
>> should be here.
>>
>> I'm still under testing, but I see unexpected behavior on A1 SoC:
>>
>> [ 1.482446] Call trace:
>> [ 1.484860] dump_backtrace+0x94/0xec
>> [ 1.488482] show_stack+0x18/0x24
>> [ 1.491754] dump_stack_lvl+0x78/0x90
>> [ 1.495377] dump_stack+0x18/0x24
>> [ 1.498654] __report_bad_irq+0x38/0xe4
>> [ 1.502453] note_interrupt+0x324/0x374
>> [ 1.506244] handle_irq_event+0x9c/0xac
>> [ 1.510039] handle_fasteoi_irq+0xa4/0x238
>> [ 1.514093] generic_handle_domain_irq+0x2c/0x44
>> [ 1.518664] gic_handle_irq+0x40/0xc4
>> [ 1.522287] call_on_irq_stack+0x24/0x4c
>> [ 1.526168] do_interrupt_handler+0x80/0x84
>> [ 1.530308] el1_interrupt+0x34/0x68
>> [ 1.533844] el1h_64_irq_handler+0x18/0x24
>> [ 1.537898] el1h_64_irq+0x64/0x68
>> [ 1.541262] default_idle_call+0x28/0x3c
>> [ 1.545143] do_idle+0x208/0x260
>> [ 1.548334] cpu_startup_entry+0x38/0x3c
>> [ 1.552220] kernel_init+0x0/0x1dc
>> [ 1.555579] start_kernel+0x5ac/0x6f0
>> [ 1.559206] __primary_switched+0x80/0x88
>
> That stack trace is not helping
> >>
>>
>> and using of regmap_write() (apparently) fixes it.
>
> Nor does that conclusion.
> > It is perfectly possible I have made mistake somewhere (I have already
> fixed one in v2), but sending incomplete report like this, with
> unargumented conclusion is just noise in the end.
>
> No, there is no situation in which `regmap_write` would solve a problem
> with `regmap_update_bits`.
>
What is the default regs' value of this reset controller? The doc that I
have doesn't clearly specifies it, but it tells that "the reset will
auto-cover to 0 by HW". However pr_info() say that there is 0xffffffff
before write (I am talking about A1).
Also we know, that reset is triggered by writing 1 to specific bit. So,
what will happen if 1 will be written where we did not intend to write it?
> Either send a full analysis of the problem you found, if you did one, or
> at least send the full dump, explaining that you don't know what is
> happening.
>
Full analysis is following:
- using regmap_update_bits() instead of writel() is incorrect because
this changes the behavior of the driver
- regmap_update_bits() should not be used here because default value of
regs isn't taken into account and (_apparently_, the doc is terse) these
regs could be updated by hw itself.
> I'll also point that I did not send g12 clock along with v1 and that a1
> support is not upstream or even reviewed. So this 'tests' obviously rely
> on an integration of v1, rfc and your custom develpment. That is hardly a
> reference at this point.
>
> This patches address g12 and sm1. If you want to contribute with tests,
> please test on these platforms. a1 will be addressed in due time.
>
Where did you find in my words any reference to not upstreamed
functionality? Support of A1 SoC is here since commit bdb369e1e98a
("reset: add support for the Meson-A1 SoC Reset Controller").
> Testing is good but making complete report (trace and environment
> details) is equally important, especially for maintainers who might not
> be following Amlogic development as much as you do.
>
>>
>>> }
>>>
>>> static int meson_reset_level(struct reset_controller_dev *rcdev,
>>> @@ -48,25 +56,13 @@ static int meson_reset_level(struct reset_controller_dev *rcdev,
>>> {
>>> struct meson_reset *data =
>>> container_of(rcdev, struct meson_reset, rcdev);
>>> - unsigned int bank = id / BITS_PER_REG;
>>> - unsigned int offset = id % BITS_PER_REG;
>>> - void __iomem *reg_addr;
>>> - unsigned long flags;
>>> - u32 reg;
>>> + unsigned int offset, bit;
>>>
>>> - reg_addr = data->reg_base + data->param->level_offset + (bank << 2);
>>> + meson_reset_offset_and_bit(data, id, &offset, &bit);
>>> + offset += data->param->level_offset;
>>>
>>> - spin_lock_irqsave(&data->lock, flags);
>>> -
>>> - reg = readl(reg_addr);
>>> - if (assert)
>>> - writel(reg & ~BIT(offset), reg_addr);
>>> - else
>>> - writel(reg | BIT(offset), reg_addr);
>>> -
>>> - spin_unlock_irqrestore(&data->lock, flags);
>>> -
>>> - return 0;
>>> + return regmap_update_bits(data->map, offset,
>>> + BIT(bit), assert ? 0 : BIT(bit));
>>> }
>>>
>>> static int meson_reset_assert(struct reset_controller_dev *rcdev,
>>> @@ -113,30 +109,42 @@ static const struct of_device_id meson_reset_dt_ids[] = {
>>> };
>>> MODULE_DEVICE_TABLE(of, meson_reset_dt_ids);
>>>
>>> +static const struct regmap_config regmap_config = {
>>> + .reg_bits = 32,
>>> + .val_bits = 32,
>>> + .reg_stride = 4,
>>> +};
>>> +
>>> static int meson_reset_probe(struct platform_device *pdev)
>>> {
>>> + struct device *dev = &pdev->dev;
>>> struct meson_reset *data;
>>> + void __iomem *base;
>>>
>>> - data = devm_kzalloc(&pdev->dev, sizeof(*data), GFP_KERNEL);
>>> + data = devm_kzalloc(dev, sizeof(*data), GFP_KERNEL);
>>> if (!data)
>>> return -ENOMEM;
>>>
>>> - data->reg_base = devm_platform_ioremap_resource(pdev, 0);
>>> - if (IS_ERR(data->reg_base))
>>> - return PTR_ERR(data->reg_base);
>>> + base = devm_platform_ioremap_resource(pdev, 0);
>>> + if (IS_ERR(base))
>>> + return PTR_ERR(base);
>>>
>>> - data->param = of_device_get_match_data(&pdev->dev);
>>> + data->param = of_device_get_match_data(dev);
>>> if (!data->param)
>>> return -ENODEV;
>>>
>>> - spin_lock_init(&data->lock);
>>> + data->map = devm_regmap_init_mmio(dev, base, ®map_config);
>>> + if (IS_ERR(data->map))
>>> + return dev_err_probe(dev, PTR_ERR(data->map),
>>> + "can't init regmap mmio region\n");
>>>
>>> data->rcdev.owner = THIS_MODULE;
>>> - data->rcdev.nr_resets = data->param->reg_count * BITS_PER_REG;
>>> + data->rcdev.nr_resets = data->param->reg_count * BITS_PER_BYTE
>>> + * regmap_config.reg_stride;
>>> data->rcdev.ops = &meson_reset_ops;
>>> - data->rcdev.of_node = pdev->dev.of_node;
>>> + data->rcdev.of_node = dev->of_node;
>>>
>>> - return devm_reset_controller_register(&pdev->dev, &data->rcdev);
>>> + return devm_reset_controller_register(dev, &data->rcdev);
>>> }
>>>
>>> static struct platform_driver meson_reset_driver = {
>
--
Best regards
Jan Dakinevich
Powered by blists - more mailing lists