linux-kernel - Re: [PATCH v4 2/8] irqchip: Add Realtek RTD1295 mux driver

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

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <37b3b5d3-b3c8-b513-f8b5-9054f32a4b53@suse.de>
Date:   Wed, 20 Nov 2019 13:12:31 +0100
From:   Andreas Färber <afaerber@...e.de>
To:     Marc Zyngier <maz@...nel.org>
Cc:     linux-realtek-soc@...ts.infradead.org,
        linux-arm-kernel@...ts.infradead.org, linux-kernel@...r.kernel.org,
        Aleix Roca Nonell <kernelrocks@...il.com>,
        James Tai <james.tai@...ltek.com>,
        Thomas Gleixner <tglx@...utronix.de>,
        Jason Cooper <jason@...edaemon.net>
Subject: Re: [PATCH v4 2/8] irqchip: Add Realtek RTD1295 mux driver

Am 20.11.19 um 11:18 schrieb Marc Zyngier:
> On 2019-11-19 23:25, Andreas Färber wrote:
>> Am 19.11.19 um 13:01 schrieb Marc Zyngier:
>>> On 2019-11-19 02:19, Andreas Färber wrote:
>>>> diff --git a/drivers/irqchip/irq-rtd1195-mux.c
>>>> b/drivers/irqchip/irq-rtd1195-mux.c
>>>> new file mode 100644
>>>> index 000000000000..e6b08438b23c
>>>> --- /dev/null
>>>> +++ b/drivers/irqchip/irq-rtd1195-mux.c
>> [...]
>>>> +static void rtd1195_mux_irq_handle(struct irq_desc *desc)
>>>> +{
>>>> +    struct rtd1195_irq_mux_data *data =
>>>> irq_desc_get_handler_data(desc);
>>>> +    struct irq_chip *chip = irq_desc_get_chip(desc);
>>>> +    u32 isr, mask;
>>>> +    int i;
>>>> +
>>>> +    chained_irq_enter(chip, desc);
>>>> +
>>>> +    isr = readl_relaxed(data->reg_isr);
>>>> +
>>>> +    while (isr) {
>>>> +        i = __ffs(isr);
>>>> +        isr &= ~BIT(i);
>>>> +
>>>> +        mask = data->info->isr_to_int_en_mask[i];
>>>> +        if (mask && !(data->scpu_int_en & mask))
>>>> +            continue;
>>>> +
>>>> +        if (!generic_handle_irq(irq_find_mapping(data->domain, i)))
>>>> +            writel_relaxed(BIT(i), data->reg_isr);
>>>
>>> What does this write do exactly? It is the same thing as a 'mask',
>>> which is pretty odd. So either:
>>>
>>> - this is not doing anything and your 'mask' callback is bogus
>>>   (otherwise you'd never have more than a single interrupt)
>>>
>>> - or this is an ACK operation, and this should be described as
>>>   such (and then fix the mask/unmask/enable/disable mess that
>>>   results from it).
>>
>> This is supposed to be an ACK, i.e. clear-1-bits operation.
> 
> If it is an ACK, model it as such, and do not open-code it.

I have found an .irq_ack callback - moving this there appears to work.

Alternatively there is an irq_eoi callback and an IRQCHIP_EOI_IF_HANDLED
flag.

It would really help me if you spelled out explicitly where you think I
should be moving code, as the documentation in irq.h is not all that
helpful in terms of when are they called and what should be done there.
In case not obvious, this is my first irqchip driver.

>>
>> The BSP had extended various drivers, such as 8250 UART, to do this ack
>> in their interrupt handler through an additional DT reg region. I tried
>> to clean that up by handling it centrally here in the irqchip driver.
>>
>>>
>>> as I can't see how the same register can be used for both purposes.
>>> You should be able to verify this experimentally, even without
>>> documentation.
>>
>> There are three registers here:
>>
>> MIS_UMSK_ISR    - MISC unmasked interrupt status register
>> MIS_ISR         - MISC   masked interrupt status register
>> MIS_SCPU_INT_EN - MISC SCPU interrupt enable register
>>
>> The latter is a regular R/W register; the former two have a write_data
>> field as BIT(0), with 1 indicating a write vs. 0 indicating clear, RAZ.
>>
>> By enabling/disabling in _SCPU_INT_EN we mask/unmask them in _ISR but
>> not in _UMSK_ISR.
>>
>> Does that shed any more light?
> 
> None whatsoever. Your mask callback doesn't make any sense, since it
> actually acks the interrupt. My gut feeling is that your enable/disable
> should really be mask/unmask.

Renaming enable -> unmask and disable -> mask works okay.

Thanks,
Andreas

-- 
SUSE Software Solutions Germany GmbH
Maxfeldstr. 5, 90409 Nürnberg, Germany
GF: Felix Imendörffer
HRB 36809 (AG Nürnberg)