Message-ID: <CA+H2tpEFQjG0fFR-D+L2KYr0uLQqYYQM8wGut7ohzLSfotV3zA@mail.gmail.com>
Date:	Tue, 20 Jan 2015 16:41:44 +0800
From:	Orson Zhai <orsonzhai@...il.com>
To:	Lyra Zhang <zhang.lyra@...il.com>
Cc:	Rob Herring <robherring2@...il.com>,
	Chunyan Zhang <chunyan.zhang@...eadtrum.com>,
	Greg Kroah-Hartman <gregkh@...uxfoundation.org>,
	Mark Rutland <mark.rutland@....com>,
	Arnd Bergmann <arnd@...db.de>,
	One Thousand Gnomes <gnomes@...rguk.ukuu.org.uk>,
	Mark Brown <broonie@...nel.org>,
	Rob Herring <robh+dt@...nel.org>,
	Pawel Moll <pawel.moll@....com>,
	Ian Campbell <ijc+devicetree@...lion.org.uk>,
	Kumar Gala <galak@...eaurora.org>,
	Will Deacon <will.deacon@....com>,
	Catalin Marinas <catalin.marinas@....com>,
	Jiri Slaby <jslaby@...e.cz>,
	Jason Cooper <jason@...edaemon.net>,
	Heiko Stübner <heiko@...ech.de>,
	Florian Vaussard <florian.vaussard@...l.ch>,
	Andrew Lunn <andrew@...n.ch>,
	Robert Richter <rrichter@...ium.com>,
	Hayato Suzuki <hytszk@...il.com>,
	Grant Likely <grant.likely@...aro.org>,
	Antony Pavlov <antonynpavlov@...il.com>,
	Joel Schopp <Joel.Schopp@....com>,
	Suravee Suthikulanit <Suravee.Suthikulpanit@....com>,
	Shawn Guo <shawn.guo@...aro.org>, lea.yan@...aro.org,
	Jorge Ramirez-Ortiz <jorge.ramirez-ortiz@...aro.org>,
	"geng.ren@...eadtrum.com" <geng.ren@...eadtrum.com>,
	"zhizhou.zhang" <zhizhou.zhang@...eadtrum.com>,
	"lanqing.liu@...eadtrum.com" <lanqing.liu@...eadtrum.com>,
	Wei Qiao (乔伟) <wei.qiao@...eadtrum.com>,
	"devicetree@...r.kernel.org" <devicetree@...r.kernel.org>,
	"linux-api@...r.kernel.org" <linux-api@...r.kernel.org>,
	"linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>,
	"linux-arm-kernel@...ts.infradead.org" 
	<linux-arm-kernel@...ts.infradead.org>,
	"linux-serial@...r.kernel.org" <linux-serial@...r.kernel.org>,
	Lee Jones <lee.jones@...aro.org>
Subject: Re: [PATCH v5 5/5] tty/serial: Add Spreadtrum sc9836-uart driver support

On Tue, Jan 20, 2015 at 3:37 PM, Lyra Zhang <zhang.lyra@...il.com> wrote:
> Hi, Rob
>
> I still have a question to be conform, specific describes below:
>
> On Mon, Jan 19, 2015 at 10:11 PM, Rob Herring <robherring2@...il.com> wrote:
>> On Mon, Jan 19, 2015 at 3:55 AM, Lyra Zhang <zhang.lyra@...il.com> wrote:
>>> On Sat, Jan 17, 2015 at 12:41 AM, Rob Herring <robherring2@...il.com> wrote:
>>>> On Fri, Jan 16, 2015 at 4:00 AM, Chunyan Zhang
>>>> <chunyan.zhang@...eadtrum.com> wrote:
>>>>> Add a full sc9836-uart driver for SC9836 SoC which is based on the
>>>>> spreadtrum sharkl64 platform.
>>>>> This driver also support earlycon.
>>>>> This patch also replaced the spaces between the macros and their
>>>>> values with the tabs in serial_core.h
>>
>> [...]
>>
>>>>> +static int __init sprd_serial_init(void)
>>>>> +{
>>>>> +       int ret = 0;
>>>>> +
>>>>> +       ret = uart_register_driver(&sprd_uart_driver);
>>>>
>>>> This can be done in probe now. Then you can use module_platform_driver().
>>>>
>>>
>>> Question:
>>> 1. there are 4 uart ports configured in dt for sprd_serial, so probe
>>> will be called 4 times, but uart_register_driver only needs to be
>>> called one time, so can we use uart_driver.state to check if
>>> uart_register_driver has already been called ?
>>
>> Yes.
>>
>>> 2. if I use module_platform_driver() instead of
>>> module_init(sprd_serial_init)  and  module_exit(sprd_serial_exit) , I
>>> must move uart_unregister_driver() which is now processed in
>>> sprd_serial_exit() to sprd_remove(), there is a similar problem with
>>> probe(), sprd_remove() will also be called 4 times, and actually it
>>> should be called only one time. How can we deal with this case?
>>
>> Look at pl01x or Samsung UART drivers which have done this conversion.
>
> Samsung UART does use module_platform_driver, but pl010/pl011 doesn't.
> In the Samsung UART driver, uart_unregister_driver is processed in
> remove(), like below:
>
> static int s3c24xx_serial_remove(struct platform_device *dev)
> {
>     struct uart_port *port = s3c24xx_dev_to_port(&dev->dev);
>
>     if (port) {
>         s3c24xx_serial_cpufreq_deregister(to_ourport(port));
>         uart_remove_one_port(&s3c24xx_uart_drv, port);
>     }
>
>     uart_unregister_driver(&s3c24xx_uart_drv);
> }
>
> if this serial has more than one ports, uart_unregister_driver() must
> be called multiple times when the device need to be removed.
> I think there may be a problem because that uart_unregister_driver()
> will do kfree(drv->state) every time when it's called.

I think it is no appropriate to call uart_unregister_driver() at first
port removing.
The drv->state buffer was shared with all uart ports.
If there are some cases that only 1 port is needed to be removed, that
will destroy all others, isn't it?

Regards,
Orson


>
> Thanks,
> Chunyan
>
>>
>>> 3. for the second question, we can check the platform_device->id, if
>>> it is equal to the index of last port (e.g. 4 for this case), then
>>> uart_unregister_driver() can be called. Does it work correctly? since
>>> for this case, we must keep the order of releasing ports.
>>
>> The id will not be the line index in the DT case. I don't think you
>> can guarantee the order either.
>>
>> It would be better to make uart_{un}register_driver deal with being
>> called multiple times so drivers don't have to deal with getting this
>> correct. I'm not sure if that is feasible though.
>>
>> Rob
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

Powered by blists - more mailing lists