linux-kernel - Re: [PATCH] drivers: tty: serial: 8250: fintek: Can enable or disable irq sharing based on isa or pci bus

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]

Date:   Fri, 13 Dec 2019 12:50:33 +0200
From:   Andy Shevchenko <andriy.shevchenko@...ux.intel.com>
To:     chengang@...ndsoft.com.cn
Cc:     gregkh@...uxfoundation.org, jslaby@...e.com, sr@...x.de,
        mika.westerberg@...ux.intel.com, yegorslists@...glemail.com,
        yuehaibing@...wei.com, haolee.swjtu@...il.com, dsterba@...e.com,
        mojha@...eaurora.org, linux-serial@...r.kernel.org,
        linux-kernel@...r.kernel.org,
        Lv Li-song <lvlisong@...ndsoft.com.cn>
Subject: Re: [PATCH] drivers: tty: serial: 8250: fintek: Can enable or
 disable irq sharing based on isa or pci bus

On Fri, Dec 13, 2019 at 01:17:17PM +0800, chengang@...ndsoft.com.cn wrote:
> From: Chen Gang <chengang@...ndsoft.com.cn>

Thanks for the patch, my comments below.

> Sorry for this patch being too late, which is for linux-next 20151127 (
> about linux 4.4-rc2).  After 4 years, much things have been changed. But
> I think it might be still valuable for some old versions. Welcome anyone
> to refact this patch for their own.

This part should go after '---' line below.

> Fintek serial ports can share irq, but they need be enabled firstly, so
> enable or disable irq sharing based on isa or pci bus. From kconfig, it

irq -> IRQ
isa -> ISA
pci -> PCI
kconfig -> Kconfig

> can be configured.
> 
> For integrated 8250 drivers, kernel always calls pnp driver, which will
> not use integrated fintek driver for ever. So let pnp driver try the

fintek -> Fintek or Fintek 8250

> other drivers firstly (e.g. fintek), if fail, try pnp driver its own.

Ditto.

...

> --- a/drivers/tty/serial/8250/8250.h
> +++ b/drivers/tty/serial/8250/8250.h
> @@ -14,6 +14,7 @@
>  #include <linux/serial_8250.h>
>  #include <linux/serial_reg.h>
>  #include <linux/dmaengine.h>

> +#include <linux/pnp.h>

The changes below doesn't require a header.
Pointers are known to the compiler. Names of data structures can be forward
declared. Moreover, these declarations may go inside respective #ifdef.

...

> +#if IS_ENABLED(CONFIG_SERIAL_8250_FINTEK_IRQ_SHARING)
> +
> +static void set_icsr(u16 base_port, u8 index)
> +{
> +	uint8_t icsr = 0;
> +
> +	outb(LDN, base_port + ADDR_PORT);
> +	outb(index, base_port + DATA_PORT);
> +	outb(ICSR, base_port + ADDR_PORT);
> +	icsr = inb(base_port + DATA_PORT);
> +

> +	if (icsr != 0xff) {

	if (icsr == 0xff)
		return;

?

> +		icsr |= IRQ_SHARING_MOD;
> +#if IS_ENABLED(CONFIG_SERIAL_8250_FINTEK_IRQ_SHARING_ISA)
> +		icsr |= ISA_IRQ_SHARING;
> +#else
> +		icsr |= PCI_IRQ_SHARING;
> +#endif
> +		outb(ICSR, base_port + ADDR_PORT);
> +		outb(icsr, base_port + DATA_PORT);
> +	}
> +}
> +
> +#endif

...

>  				aux |= inb(addr[i] + DATA_PORT) << 8;
>  				if (aux != io_address)
>  					continue;

> -

What the point?

> +#if IS_ENABLED(CONFIG_SERIAL_8250_FINTEK_IRQ_SHARING)
> +				set_icsr(addr[i], k);
> +#endif
>  				fintek_8250_exit_key(addr[i]);
>  				*key = keys[j];
>  				*index = k;
> @@ -179,53 +212,6 @@ static int fintek_8250_base_port(u16 io_address, u8 *key, u8 *index)
>  	return -ENODEV;
>  }
>  
> -static int
> -fintek_8250_probe(struct pnp_dev *dev, const struct pnp_device_id *dev_id)

Why did you move this function?
It's now not only hard to follow what has been changed, and to review.

> --- a/drivers/tty/serial/8250/8250_pnp.c
> +++ b/drivers/tty/serial/8250/8250_pnp.c
> @@ -438,8 +438,13 @@ static int
>  serial_pnp_probe(struct pnp_dev *dev, const struct pnp_device_id *dev_id)
>  {
>  	struct uart_8250_port uart, *port;
> -	int ret, line, flags = dev_id->driver_data;
> +	int ret, line, flags;
>  

> +#if IS_BUILTIN(CONFIG_SERIAL_8250_FINTEK)
> +	if (!fintek_8250_probe(dev, dev_id))
> +		return 0;
> +#endif
> +	flags = dev_id->driver_data;

Oh, I don't like this.
It needs a bit more refactoring done first.

The idea that we are not going to pollute generic driver(s) with quirks anymore
(only when it's really unavoidable).


-- 
With Best Regards,
Andy Shevchenko