From mboxrd@z Thu Jan 1 00:00:00 1970 From: Lee Schermerhorn Subject: [BUG?] linux-next/mmotm: boot hang on HP ia64 servers Date: Mon, 06 Apr 2009 13:55:42 -0400 Message-ID: <1239040542.14812.210.camel@lts-notebook> Mime-Version: 1.0 Content-Type: text/plain Content-Transfer-Encoding: 7bit Return-path: Received: from g5t0006.atlanta.hp.com ([15.192.0.43]:5808 "EHLO g5t0006.atlanta.hp.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1752172AbZDFRzs (ORCPT ); Mon, 6 Apr 2009 13:55:48 -0400 Sender: linux-next-owner@vger.kernel.org List-ID: To: Andrew Morton , Alan Cox , Clark Williams Cc: linux-next@vger.kernel.org, Bob Picco , Bjorn Helgaas Apparently caused by: commit c5b01e769cee440772d6a8cd1c71c0acf37451fc -- "disallow SERIAL_8250_PNP with SERIAL_8250_CONSOLE". I haven't been able to boot the mmotm tree on one of my test servers for weeks now, but have been too busy with other things to investigate. On these servers, this serial line connects the main server to the "management processor" which is accessed via the network. The console shows up as ttyS0 in the booted kernel. Symptom of the hang was this message on the console: turn off boot console uart8250 and then nothing. I don't recall seeing that message on any kernel that booted successfully. But I tend to boot with "debug" and have lots of hardware to report, so it might have gotten lost. I thought maybe the system would continue to come up "blind" [or should that be "mute"?], but seems it never did. I also hoped that the problem would "go away" in later mmotms but, again, no luck :). Finally got a change to look in detail at linux-next in mmotm 03/23 [and 04/05], and noticed the commit above. Comparing configs for kernels that booted [!mmotm] and those that did, I saw that "CONFIG_SERIAL_8250_PNP=y" was missing from the ones that don't boot. Today, I reverted this patch from the latest [04/05] mmotm, added back that config line and it now boots. The patch description indicates "hard to fix issues" when both SERIAL_8250_CONSOLE and SERIAL_8250_PNP are both set. However, we can't seem to boot with out both. For what it's worth, I've included the patch that reverts the [problematic portion of the] commit. Regards, Lee P.S., Bjorn says "I told you so" :) ============================== HP sxX000 ia64 servers hang during boot unless SERIAL_8250_PNP is configured. This patch reverts the patch that prevents both SERIAL_8250_CONSOLE and SERIAL_8250_PNP from being configured at the same time. Not sure this is the right thing to do. Signed-off-by: Lee Schermerhorn drivers/serial/Kconfig | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) Index: linux-2.6.29-mmotm-090405-2145/drivers/serial/Kconfig =================================================================== --- linux-2.6.29-mmotm-090405-2145.orig/drivers/serial/Kconfig 2009-04-06 13:09:54.000000000 -0400 +++ linux-2.6.29-mmotm-090405-2145/drivers/serial/Kconfig 2009-04-06 13:09:55.000000000 -0400 @@ -91,7 +91,7 @@ config SERIAL_8250_PCI config SERIAL_8250_PNP tristate "8250/16550 PNP device support" if EMBEDDED - depends on SERIAL_8250 && PNP && ! SERIAL_8250_CONSOLE + depends on SERIAL_8250 && PNP default SERIAL_8250 help This builds standard PNP serial support. You may be able to