Re: No serial since kernel 3.8

[Greg KH <gregkh@linuxfoundation.org>]

On Thu, Apr 18, 2013 at 01:15:27PM +0200, Stephan von Krawczynski wrote:
> On Wed, 17 Apr 2013 20:55:04 +0200
> "Rafael J. Wysocki" <rjw@sisk.pl> wrote:
> 
> > On Wednesday, April 17, 2013 11:38:30 AM Bjorn Helgaas wrote:
> > > [+cc Rafael & linux-acpi]
> > 
> > Thanks.
> > 
> > > On Tue, Apr 16, 2013 at 10:14 AM, Stephan von Krawczynski
> > > <skraw@ithnet.com> wrote:
> > > > On Tue, 16 Apr 2013 10:59:53 -0400
> > > > Josh Boyer <jwboyer@gmail.com> wrote:
> > > >
> > > >> On Tue, Apr 16, 2013 at 04:55:25PM +0200, richard -rw- weinberger wrote:
> > > >> >On Tue, Apr 16, 2013 at 4:20 PM, Stephan von Krawczynski
> > > >> ><skraw@ithnet.com> wrote:
> > > >> >> Hello,
> > > >> >>
> > > >> >> can some kind soul please explain how I can get my serial port back under
> > > >> >> kernel 3.8.X. Earlier kernels showed:
> > > >> >>
> > > >> >> Jan 10 13:57:18 mybox kernel: [    0.712829] serial8250: ttyS0 at I/O 0x3f8
> > > >> >> (irq = 4) is a 16550A
> > > >> >>
> > > >> >> But 3.8.4:
> > > >> >>
> > > >> >> Mar 26 10:39:14 admin kernel: [    0.603647] serial 00:0a: disabled
> > > >> >>
> > > >> >> Only kernel is replaced, no change in distro
> > > >> >> Is there some kernel boot parameter needed now?
> > > >> >> I tried 3.4.X and it works, even 3.7.X works...
> > > >> >
> > > >> >So, 3.8.3 worked and 3.8.4 introduced a regression?
> > > >> >If so please confirm.
> > > >> >
> > > >> >CC'ing gregkh, just in case...
> > > >>
> > > >> We had a report of this in Fedora that was fixed with 3.8.7.  An upstream
> > > >> change had been brought back to the 3.8.x series, and it was subsequently
> > > >> reverted.  3.8.7 picked up the revert.
> > > >>
> > > >> josh
> > > >
> > > > 3.8.7 does not work either. I just checked it. I tried shared and not shared IRQ:
> > > >
> > > > Apr 16 17:47:05 mybox kernel: [    0.604826] Serial: 8250/16550 driver, 8 ports, IRQ sharing disabled
> > > >
> > > > and another test:
> > > >
> > > > Apr 16 18:07:08 mybox kernel: [    0.601203] Serial: 8250/16550 driver, 8 ports, IRQ sharing enabled
> > > >
> > > > I must say that 3.8.7 does not even tell the "serial 00:0a: disabled" any more.
> > > > This is no special board, it is a pretty mainstream asus xeon board with onboard serial.
> > > > Loading 3.4.24 (just an example) makes the port work immediately.
> > > > Anything else I can test?
> > > 
> > > 3.7.6 works, and 3.8.3, 3.8.4, 3.8.7 all fail.  So it looks like a
> > > regression between 3.7.6 and 3.8.3.
> > > 
> > > The "serial 00:0a: disabled" message is from PNP, so I added Rafael
> > > and linux-acpi.  Complete dmesg logs from 3.7.6 and 3.8.4 would be a
> > > place to start.  Also maybe an acpidump, since the 00:0a PNP device
> > > probably came from PNPACPI.  You could attach all these to a
> > > https://bugzilla.kernel.org report if you want a place to stash them.
> > > 
> > > Josh, do you have a pointer (URL) to the Fedora issue so we can rule
> > > that in or out?
> > 
> > Stephan, below is a list of commits touching the PNP code between 3.7 and 3.8.
> > 
> > Can you please check if any of them has triggered the problem you're seeing?
> > 
> > Rafael
> > 
> > 
> > 13cde3b PNP: Handle IORESOURCE_BITS in resource allocation
> > c937766 PNP: Simplify setting of resources
> > cdc87c5 pnpacpi: fix incorrect TEST_ALPHA() test
> > a6b5e88 ACPI / PNP: Do not crash due to stale pointer use during system resume
> > 38de279 pnpbios: remove CONFIG_HOTPLUG ifdefs
> > 2905875 ACPI / PNP: skip ACPI device nodes associated with physical nodes already
> > 046d9ce ACPI: Move device resources interpretation code from PNP to ACPI core
> > 
> > -- 
> > I speak only for myself.
> > Rafael J. Wysocki, Intel Open Source Technology Center.
> 
> Hello all,
> 
> I checked 3.8.8 today and it _does_ work.
> Shall we burn some time to find out why?
> The only significant output difference between working and not working we could find is:
> 
> not working:
> Apr  2 17:28:23 mybox kernel: [    0.603270] serial 00:0a: disabled
> 
> working (3.8.8):
> Apr 18 12:58:50 mybox kernel: [    2.855929] 00:0a: ttyS0 at I/O 0x3f8 (irq = 4) is a 16550A

If you are curious, you can run 'git bisect' to figure out the fix, but
if you look at the commits in these 3.8.x releases, I think it will be
pretty obvious as to what specific patch fixed it :)

Thanks for letting us know,

greg k-h