[PATCH] TTY: hvc_dcc: probe for a JTAG connection before registering

Rob Herring robherring2 at gmail.com
Mon Sep 23 17:35:51 EDT 2013


On 09/23/2013 04:10 PM, Tim Gardner wrote:
> On 09/23/2013 01:30 PM, Rob Herring wrote:
>> From: Rob Herring <rob.herring at calxeda.com>
>>
>> Enabling the ARM DCC console and using without a JTAG connection will
>> simply hang the system. Since distros like to turn on all options, this
>> is a reoccurring problem to debug. We can do better by checking if
>> anything is attached and handling characters. There is no way to probe
>> this, so send a newline and check that it is handled.
>>
>> Cc: Paolo Pisati <paolo.pisati at canonical.com>
>> Cc: Tim Gardner <tim.gardner at canonical.com>
>> Cc: Greg Kroah-Hartman <gregkh at linuxfoundation.org>
>> Cc: Jiri Slaby <jslaby at suse.cz>
>> Signed-off-by: Rob Herring <rob.herring at calxeda.com>
>> ---
>>  drivers/tty/hvc/hvc_dcc.c | 21 +++++++++++++++++++++
>>  1 file changed, 21 insertions(+)
>>
>> diff --git a/drivers/tty/hvc/hvc_dcc.c b/drivers/tty/hvc/hvc_dcc.c
>> index 44fbeba..44c8aff 100644
>> --- a/drivers/tty/hvc/hvc_dcc.c
>> +++ b/drivers/tty/hvc/hvc_dcc.c
>> @@ -86,6 +86,21 @@ static int hvc_dcc_get_chars(uint32_t vt, char *buf, int count)
>>  	return i;
>>  }
>>  
>> +static bool hvc_dcc_check(void)
>> +{
>> +	unsigned long time = jiffies;
>> +
>> +	/* Write a test character to check if it is handled */
>> +	__dcc_putchar('\n');
>> +
>> +	while (jiffies < time + (HZ / 10)) {
>> +		if (!(__dcc_getstatus() & DCC_STATUS_TX))
>> +			return true;
>> +	}
> 
> I think you should use one of the jiffies time comparison functions, e.g.,
> 
> 	time = jiffies + (HZ / 10);
> 	while (time_before(jiffies,time) {
> 		...
> 	}

Yes.

> 
> What is the possibility that __dcc_getstatus() will return garbage if no
> JTAG is connected ? In that event you're likely to get a false positive
> on DCC_STATUS_TX.

That probably would not be architecturally compliant. The status bits
are cleared by JTAG reading the read side of the register. If the bit
stays set, we'll return false. If the bit is always clear, then we
wouldn't hang in the first place and should be fine. If you get random
values, then I'm sorry and good luck. You are still no worse off than
before this patch.

Rob




More information about the linux-arm-kernel mailing list