linux-kernel - Re: [PATCH v1 2/5] tty: serial: meson: redesign the module to platform

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: <20230704151920.ryjab2kcux5avwfi@CAB-WSD-L081021>
Date:   Tue, 4 Jul 2023 18:19:20 +0300
From:   Dmitry Rokosov <ddrokosov@...rdevices.ru>
To:     <neil.armstrong@...aro.org>
CC:     <gregkh@...uxfoundation.org>, <robh+dt@...nel.org>,
        <krzysztof.kozlowski+dt@...aro.org>, <conor+dt@...nel.org>,
        <jbrunet@...libre.com>, <jirislaby@...nel.org>,
        <khilman@...libre.com>, <martin.blumenstingl@...glemail.com>,
        <kelvin.zhang@...ogic.com>, <xianwei.zhao@...ogic.com>,
        <kernel@...rdevices.ru>, <rockosov@...il.com>,
        <linux-amlogic@...ts.infradead.org>,
        <linux-serial@...r.kernel.org>, <devicetree@...r.kernel.org>,
        <linux-kernel@...r.kernel.org>,
        <linux-arm-kernel@...ts.infradead.org>
Subject: Re: [PATCH v1 2/5] tty: serial: meson: redesign the module to
 platform_driver

On Tue, Jul 04, 2023 at 04:46:40PM +0200, neil.armstrong@...aro.org wrote:
> On 04/07/2023 15:59, Dmitry Rokosov wrote:
> > Actually, the meson_uart module is already a platform_driver, but it is
> > currently registered manually and the uart core registration is run
> > outside the probe() scope, which results in some restrictions. For
> > instance, it is not possible to communicate with the OF subsystem
> > because it requires an initialized device object.
> > 
> > To address this issue, apply module_platform_driver() instead of direct
> > module init/exit routines. Additionally, move uart_register_driver() to
> > the driver probe(), and destroy manual console registration because it's
> > already run in the uart_register_driver() flow.
> > 
> > Signed-off-by: Dmitry Rokosov <ddrokosov@...rdevices.ru>
> > ---
> >   drivers/tty/serial/meson_uart.c | 46 +++++++--------------------------
> >   1 file changed, 10 insertions(+), 36 deletions(-)
> > 
> > diff --git a/drivers/tty/serial/meson_uart.c b/drivers/tty/serial/meson_uart.c
> > index 169f028956ae..87c0eb5f2dba 100644
> > --- a/drivers/tty/serial/meson_uart.c
> > +++ b/drivers/tty/serial/meson_uart.c
> > @@ -621,12 +621,6 @@ static struct console meson_serial_console = {
> >   	.data		= &meson_uart_driver,
> >   };
> > -static int __init meson_serial_console_init(void)
> > -{
> > -	register_console(&meson_serial_console);
> > -	return 0;
> > -}
> > -
> >   static void meson_serial_early_console_write(struct console *co,
> >   					     const char *s,
> >   					     u_int count)
> > @@ -652,9 +646,6 @@ OF_EARLYCON_DECLARE(meson, "amlogic,meson-ao-uart",
> >   #define MESON_SERIAL_CONSOLE	(&meson_serial_console)
> >   #else
> > -static int __init meson_serial_console_init(void) {
> > -	return 0;
> > -}
> >   #define MESON_SERIAL_CONSOLE	NULL
> >   #endif
> > @@ -738,6 +729,13 @@ static int meson_uart_probe(struct platform_device *pdev)
> >   	if (ret)
> >   		return ret;
> > +	if (!meson_uart_driver.state) {
> > +		ret = uart_register_driver(&meson_uart_driver);
> > +		if (ret)
> > +			return dev_err_probe(&pdev->dev, ret,
> > +				      "failed to register meson-uart driver\n");
> > +	}
> 
> PL010 protects this in a mutex, and I think you should do the same otherwise
> if multiple uart probes at the same it will do weird stuff.
> 

Looks like that not all drivers protect this location with a specialized
mutex object. Firstly, I think it's important to verify parallel probe()
calling and implementing mutex protection at the platform core level.
For example, I've faced with the same problem during regmap mutex based
protection.

> > +
> >   	port->iotype = UPIO_MEM;
> >   	port->mapbase = res_mem->start;
> >   	port->mapsize = resource_size(res_mem);
> > @@ -776,6 +774,8 @@ static int meson_uart_remove(struct platform_device *pdev)
> >   	uart_remove_one_port(&meson_uart_driver, port);
> >   	meson_ports[pdev->id] = NULL;
> > +	uart_unregister_driver(&meson_uart_driver);
> > +
> 
> This is dangerous, it will remove the driver even if some uart are still attached to it.
> 
> You should probably do like in pl010_remove() and remove only if the last one is removed.
> 

Indeed... multiple ports can be registered...

> >   	return 0;
> >   }
> > @@ -809,33 +809,7 @@ static  struct platform_driver meson_uart_platform_driver = {
> >   	},
> >   };
> > -static int __init meson_uart_init(void)
> > -{
> > -	int ret;
> > -
> > -	ret = meson_serial_console_init();
> > -	if (ret)
> > -		return ret;
> > -	
> > -	ret = uart_register_driver(&meson_uart_driver);
> > -	if (ret)
> > -		return ret;
> > -
> > -	ret = platform_driver_register(&meson_uart_platform_driver);
> > -	if (ret)
> > -		uart_unregister_driver(&meson_uart_driver);
> > -
> > -	return ret;
> > -}
> > -
> > -static void __exit meson_uart_exit(void)
> > -{
> > -	platform_driver_unregister(&meson_uart_platform_driver);
> > -	uart_unregister_driver(&meson_uart_driver);
> > -}
> > -
> > -module_init(meson_uart_init);
> > -module_exit(meson_uart_exit);
> > +module_platform_driver(meson_uart_platform_driver);
> 
> Only pl010 uses this scheme, and I don't know why... if it works then it's ok for me.

>From my point of view, the "scheme" is using uart driver registration
from the probe() routine. Many drivers are based on such approach:
samsung-tty, timbuart, sprd, max3100, etc. Some of them are platform
drivers as well.

> >   MODULE_AUTHOR("Carlo Caione <carlo@...one.org>");
> >   MODULE_DESCRIPTION("Amlogic Meson serial port driver");
> 

-- 
Thank you,
Dmitry