linux-kernel - Re: [PATCH v3] serial: 8250_uniphier: add UniPhier serial 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 for Android: free password hash cracker in your pocket

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

Date:	Tue, 19 May 2015 20:18:09 +0900
From:	Masahiro Yamada <yamada.masahiro@...ionext.com>
To:	One Thousand Gnomes <gnomes@...rguk.ukuu.org.uk>
Cc:	Matthias Brugger <matthias.bgg@...il.com>,
	Peter Hurley <peter@...leysoftware.com>,
	John Crispin <blogic@...nwrt.org>,
	Tony Lindgren <tony@...mide.com>,
	Greg Kroah-Hartman <gregkh@...uxfoundation.org>,
	Joachim Eastwood <manabian@...il.com>,
	"linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>,
	linux-serial@...r.kernel.org,
	Ricardo Ribalda Delgado <ricardo.ribalda@...il.com>,
	Andy Shevchenko <andriy.shevchenko@...ux.intel.com>,
	Jiri Slaby <jslaby@...e.cz>,
	Sebastian Andrzej Siewior <bigeasy@...utronix.de>,
	<"linux-arm-kernel@...ts.infradead.org\" <linux-arm-kernel@...ts.infradead.org>"@lxorguk.ukuu.org.uk>
Subject: Re: [PATCH v3] serial: 8250_uniphier: add UniPhier serial driver

Hi Alan,


2015-05-19 18:48 GMT+09:00 One Thousand Gnomes <gnomes@...rguk.ukuu.org.uk>:
>> I intentionally use deeper indentation for *_SHIFT
>> because I want to clearly show  UNIPHIER_UART_LCR_SHIFT
>> belongs to UNIPHIER_UART_LCR_MCR register.
>
> Seems sensible to do it that way to me and a lot of other bits of the
> kernel do.
>
> The only other question I have is about the unipher_serial_out. If I am
> writing a "special" register then the sequence becomes
>
> - read 32bits
> - modify
> - write 32bits
>
> That means that it's no longer atomic safe as the kernel expects.
> Checking the users FCR seems safe, LCR is probably safe and MCR likewise
> so I don't see a problem but I think it's worth noting in case anyone
> else does.

Uh, I missed this.
I am a bit afraid what if LCR and MCR are updated at the same time.

Is it better to add mutex for writing special case registers?

    if (normal) {
            writel(value, p->membase + offset);
    } else {
            /* special case: two registers share the same address. */
            u32 tmp = readl(p->membase + offset);
            struct uniphier8250_priv *priv = p->private_data;

            mutex_lock(&priv->atomic_write_lock);
            tmp &= ~(0xff << valshift);
            tmp |= value << valshift;
            writel(tmp, p->membase + offset);
            mutex_unlock(&priv->atomic_write_lock);
    }


If it is OK, I can fix it in v4.




> Finally can you add a comment in serial_in and serial_out where one
> switch case drops through into the next so its obvious to anyone looking
> at Coverity and other analyser output that this drop through was
> intentional ?

I thought about it, too.

My previous version was as follows:


+#define UNIPHIER_UART_CHAR_FCR 3
+#define     UNIPHIER_UART_CHAR_SHIFT   8       /* Character Register */
+#define     UNIPHIER_UART_FCR_SHIFT    0       /* FIFO Control Register */
+#define UNIPHIER_UART_LCR_MCR  4
+#define     UNIPHIER_UART_LCR_SHIFT    8       /* Line Control Register */
+#define     UNIPHIER_UART_MCR_SHIFT    0       /* Modem Control Register */
+#define UNIPHIER_UART_DLR      9               /* Divisor Latch Register */

[snip]

+static void uniphier_serial_out(struct uart_port *p, int offset, int value)
+{
+       int valshift = 0;
+       bool normal = false;
+
+       switch (offset) {
+       case UART_FCR:
+               offset = UNIPHIER_UART_CHAR_FCR;
+               valshift = UNIPHIER_UART_FCR_SHIFT;
+               break;
+       case UART_LCR:
+               offset = UNIPHIER_UART_LCR_MCR;
+               valshift = UNIPHIER_UART_LCR_SHIFT;
+               /* Divisor latch access bit does not exist. */
+               value &= ~(UART_LCR_DLAB << valshift);
+               break;
+       case UART_MCR:
+               offset = UNIPHIER_UART_LCR_MCR;
+               valshift = UNIPHIER_UART_MCR_SHIFT;
+               break;
+       default:
+               normal = true;
+               break;
+       }



I thought it was clear to anyone although it was a bit redundant and
Matthias was opposed to it.

I personally prefer clear code to tricky code that requires comments.


-- 
Best Regards
Masahiro Yamada
--
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/