Subject: Re: console issue since 3.6, console=ttyS1 hangs
To: Sean Young <sean@mess.org>
References: <20161017144951.GA235535@stormcage.americas.sgi.com>
 <20161017151906.GA19082@gofer.mess.org>
 <20161017164140.GA35950@stormcage.americas.sgi.com>
 <20161018164004.GA159046@stormcage.americas.sgi.com>
 <20161018180518.GA28700@gofer.mess.org>
 <20161018192930.GA207959@stormcage.americas.sgi.com>
 <20161019090719.GA1764@gofer.mess.org>
CC: <linux-kernel@vger.kernel.org>, <linux-serial@vger.kernel.org>,
        <gregkh@linuxfoundation.org>, <alan@linux.intel.com>
From: Nathan Zimmer <nzimmer@sgi.com>
Message-ID: <29b311e3-a26e-81de-82cf-e0a202db6d96@sgi.com>
Date: Wed, 19 Oct 2016 17:13:41 -0500
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:45.0) Gecko/20100101
 Thunderbird/45.3.0
MIME-Version: 1.0
In-Reply-To: <20161019090719.GA1764@gofer.mess.org>
Content-Type: text/plain; charset="windows-1252"; format=flowed
Content-Transfer-Encoding: 7bit
Sender: linux-kernel-owner@vger.kernel.org
Content-Length: 2917
Lines: 76


On 10/19/2016 04:07 AM, Sean Young wrote:
> On Tue, Oct 18, 2016 at 02:29:30PM -0500, Nathan Zimmer wrote:
>> On Tue, Oct 18, 2016 at 07:05:18PM +0100, Sean Young wrote:
>>> On Tue, Oct 18, 2016 at 11:40:04AM -0500, Nathan Zimmer wrote:
>>>> 3.7.0
>>>> cat /sys/bus/pnp/drivers/serial/*/resources
>>>> state = active
>>>> io 0x2f8-0x2ff
>>>> irq 12
>>>> dma disabled
>>>>
>>>> 3.6.0
>>>> :~ # cat /sys/bus/pnp/drivers/serial/*/resources
>>>> cat: /sys/bus/pnp/drivers/serial/*/resources: No such file or directory
>>> irq 12 for ttyS1? That should be irq 3. The bios is putting bogus information
>>> in pnp. Maybe there is rubbish in the bios setup or maybe it's fixed in a
>>> newer bios update.
>>>
>>> So before this change, the kernel would assume irq 3. After this change,
>>> the kernel first uses the information in pnp to see where the serial
>>> port is. It gets told that it's irq 12 and presumably it runs into all
>>> sorts of problems then. If memory serves that's the irq for the ps/2 mouse.
>>>
>>> The interesting bit is in 3.6.0:
>>>
>>> setserial
>>> /dev/ttyS1, UART: 16550A, Port: 0x02f8, IRQ: 3
>>>
>>> becomes in 3.7.0:
>>>
>>> setserial
>>> /dev/ttyS1, UART: 16550A, Port: 0x02f8, IRQ: 12
>>>
>>> You should be able to set the right irq with setserial, but obviously
>>> that doesn't help you if it fails in early boot. It's not immediately
>>> obvious to me what can be done in the kernel for this. Maybe the dmesg
>>> output could inspire, thanks.
>> Yea the changing irq seemed weird to me too but I couldn't manage a guess to why.
>>
>> Here are the dmesgs.
> So with 3.6.0:
>
>> [    2.079980] Serial: 8250/16550 driver, 8 ports, IRQ sharing disabled
>> [    2.100887] serial8250: ttyS1 at I/O 0x2f8 (irq = 3) is a 16550A
>> [    2.101715] serial 00:04: unable to assign resources
>> [    2.102174] serial: probe of 00:04 failed with error -16
> The pnp probe fails for some reason. I don't understand why.
>
> With 3.7.0:
>
>> [    2.062700] Serial: 8250/16550 driver, 8 ports, IRQ sharing disabled
>> [    2.063250] serial 00:04: [io  0x02f8-0x02ff]
>> [    2.063875] serial 00:04: [irq 12]
>> [    2.064345] serial 00:04: [dma 18446744073709551615 disabled]
>> [    2.065540] serial 00:04: activated
>> [    2.086442] 00:04: ttyS1 at I/O 0x2f8 (irq = 12) is a 16550A
> Now the pnp probe succeeds (with broken irq from pnp).
>
> Can you please check if there is a wrong irq configured in the bios setup
> or if there is a bios update available? I don't know why this worked in
> the first place.
>
>
> Sean

Apparently this is the latest bios available for these nodes.
Also in the bios setup screens I don't see anything for changing irq 
numbers for serial console.
But this is a cluster so sometimes thing get hidden to keep everything 
uniform as possible.

If you want to point me to the pnp probe code you would be suspicious of 
I can try to debug and see what is going there.