From mboxrd@z Thu Jan 1 00:00:00 1970 From: Alan Jenkins Subject: Re: ACPI: EC: Fix logspam in "GPE storm avoidance" code Date: Tue, 28 Oct 2008 20:18:07 +0000 Message-ID: <4907737F.6060507@tuffmail.co.uk> References: <49071C33.7030308@tuffmail.co.uk> <4907590E.4070104@suse.de> Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit Return-path: Received: from nf-out-0910.google.com ([64.233.182.185]:61127 "EHLO nf-out-0910.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751897AbYJ1UQm (ORCPT ); Tue, 28 Oct 2008 16:16:42 -0400 Received: by nf-out-0910.google.com with SMTP id d3so1065818nfc.21 for ; Tue, 28 Oct 2008 13:16:38 -0700 (PDT) In-Reply-To: <4907590E.4070104@suse.de> Sender: linux-acpi-owner@vger.kernel.org List-Id: linux-acpi@vger.kernel.org To: Alexey Starikovskiy Cc: linux acpi Alexey Starikovskiy wrote: > Alan Jenkins wrote: >> >> "plenty of line "ACPI: EC: non-query interrupt received, >> switching to interrupt mode" in dmesg" > Probably, it is better to make pr_debug(). Yes, it seems useful to have when all the DEBUG output is enabled. In that case, the initial message about starting in poll mode should also be switched to pr_debug(). Otherwise, the change will give the impression that the EC never switches to interrupt mode. That just leaves the message "missing confirmations, switching to polling mode". That should only happen once per boot, or after a suspend/resume cycle, right? Because it sets NO_GPE, which inhibits GPE_MODE and is only reset on resume. So that can be preserved, and it's probably a useful message to have. I'll leave it to you to submit a more appropriate patch. >> >> GPE storm avoidance involves disabling the EC GPE during transactions. >> Unfortunately we forget to re-enable EC_FLAGS_GPE_MODE afterwards. > "We" don't forget it, "we" just use same mechanism as driver init. > Thus if > for some reason interrupts do not start to arrive after we issue > enable_gpe(), > we will not fall to timeouts. Interesting. I agree that sort of caution is meritted for the EC driver! If that's by design, then great. I will have a look at the code again for my own peace of mind; I thought I understood the GPE storm avoidance but now I'm a bit fuzzy on the details. We apologize for the overuse of the third person. After staring at the code for so long I have to identify with it to some extent, despite not having actually written any of it. >> >> This happens to work because we re-enable it anyway when we detect >> the first confirmation interrupt of the next transaction. However, >> this causes lots of bogus "switching to interrupt mode" messages. > they are not "bogus", it is really switching the mode... Bad words. It seems bogus because the message changes meaning - for me as _user_. It used to be a gentle reassurance that my computer was booting or resuming it's way towards normal working state. Now it has changed to a weaker meaning, because it can also mean that a new transaction started in GPE_STORM mode. In the technical and literal sense the meaning is unchanged. If the message is hidden under DEBUG, that is the only meaning that will matter :). Thanks Alan