linux-kernel - Re: [PATCH] ARM: at91: debug: use DEBUG_UART

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:	Wed, 8 Jun 2016 10:18:10 +0200
From:	Nicolas Ferre <nicolas.ferre@...el.com>
To:	Alexandre Belloni <alexandre.belloni@...e-electrons.com>,
	Boris Brezillon <boris.brezillon@...e-electrons.com>
CC:	<linux-arm-kernel@...ts.infradead.org>,
	Russell King <linux@...linux.org.uk>,
	Arnd Bergmann <arnd@...db.de>, <linux-kernel@...r.kernel.org>
Subject: Re: [PATCH] ARM: at91: debug: use DEBUG_UART_VIRT

Le 08/06/2016 09:17, Nicolas Ferre a écrit :
> Le 07/06/2016 18:23, Alexandre Belloni a écrit :
>> On 07/06/2016 at 17:48:21 +0200, Nicolas Ferre wrote :
>>> Le 07/06/2016 17:24, Alexandre Belloni a écrit :
>>>> AT91 still uses an offset (0x0100 0000) from the physical address to map
>>>> the debug UART. This is unfortunate as for some platforms (sama5d3 and
>>>> earlier), it ends up in the PCI zone and PCI is enabled in multi_v7.
>>>> Switch to DEBUG_UART_VIRT to solve that.
>>>>
>>>> Tested on sama5d3 and 9g20.
>>>>
>>>> Signed-off-by: Alexandre Belloni <alexandre.belloni@...e-electrons.com>
>>>
>>> People using their old defconfigs must pay attention to this change...
>>> but it's true that it's a debug configuration anyway...
>>>
>>
>> I really doubt people are letting DEBUG_LL enabled in a production
>> kernel...
>>
>>>> diff --git a/arch/arm/include/debug/at91.S b/arch/arm/include/debug/at91.S
>>>> index d4ae3b8e2426..0098401e5aeb 100644
>>>> --- a/arch/arm/include/debug/at91.S
>>>> +++ b/arch/arm/include/debug/at91.S
>>>> @@ -9,14 +9,6 @@
>>>>   *
>>>>  */
>>>>  
>>>> -#ifdef CONFIG_MMU
>>>> -#define AT91_IO_P2V(x) ((x) - 0x01000000)
>>>> -#else
>>>> -#define AT91_IO_P2V(x) (x)
>>>> -#endif
>>>> -
>>>> -#define AT91_DEBUG_UART_VIRT AT91_IO_P2V(CONFIG_DEBUG_UART_PHYS)
>>>> -
>>>>  #define AT91_DBGU_SR		(0x14)	/* Status Register */
>>>>  #define AT91_DBGU_THR		(0x1c)	/* Transmitter Holding Register */
>>>>  #define AT91_DBGU_TXRDY		(1 << 1)	/* Transmitter Ready */
>>>> @@ -24,7 +16,7 @@
>>>>  
>>>>  	.macro	addruart, rp, rv, tmp
>>>>  	ldr	\rp, =CONFIG_DEBUG_UART_PHYS		@ System peripherals (phys address)
>>>> -	ldr	\rv, =AT91_DEBUG_UART_VIRT		@ System peripherals (virt address)
>>>> +	ldr	\rv, =CONFIG_DEBUG_UART_VIRT		@ System peripherals (virt address)
>>>
>>> Shouldn't we protect the use of this defined value with some
>>> #warning "Beware the value CONFIG_DEBUG_UART_VIRT haven't been defined:
>>> is it intentional"
>>> or even #error?
>>>
>>> or something like that?
>>>
>>
>> Well, those that are using this feature are supposed to know what they
>> are doing. There is now protection for CONFIG_DEBUG_UART_PHYS anyway.
> 
> So, I know that developers are sensible people and know what they're
> doing but still... what about having a protection for both anyway...
> instead of silently crashing at runtime?

After discussion with you on IRC, it seems to be not so easy to do and
can harm the multi_v7 early debug facilities.

So, for this patch:
Acked-by: Nicolas Ferre <nicolas.ferre@...el.com>

It's true anyway that putting the proper values for these config options
is not so easy (reading the help for CONFIG_DEBUG_AT91_UART)... And
enhancing this situation if far beyond the scope of this patch.

Thanks. Bye,
-- 
Nicolas Ferre