From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from ozlabs.org (ozlabs.org [IPv6:2401:3900:2:1::2]) (using TLSv1.2 with cipher AECDH-AES256-SHA (256/256 bits)) (No client certificate requested) by lists.ozlabs.org (Postfix) with ESMTPS id 3qmZVq0LStzDq62 for ; Fri, 15 Apr 2016 21:06:59 +1000 (AEST) In-Reply-To: <20160407215811.8404.6733.stgit@hbathini.in.ibm.com> To: Hari Bathini , linuxppc-dev From: Michael Ellerman Cc: Michael Neuling , Mahesh J Salgaonkar , Paul Mackerras Subject: Re: [v4, 2/3] ppc64/book3s: make some room for common interrupt vector code Message-Id: <3qmZVp4lgdz9t5C@ozlabs.org> Date: Fri, 15 Apr 2016 21:06:58 +1000 (AEST) List-Id: Linux on PowerPC Developers Mail List List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Hi Hari, Thanks for persisting with this. On Thu, 2016-07-04 at 21:58:50 UTC, Hari Bathini wrote: > With the previous patch, we choke out whatever little space is left > below 0x7000 (FWNMI hard block) while there is a hole of ~1400 bytes > below __end_interrupts marker when CONFIG_CBE_RAS is disabled. > Considering CONFIG_CBE_RAS is not enabled by default for BOOK3S, this > is not a desirable scenario especially when we have to worry about > each additional instruction that goes below 0x7000. > > Memory region from 0x1800 to 0x4000 is dedicated for common interrupt > vector code. Also, we never hit an interrupt below 0x300 when IR=DR=1 > implying memory region between 0x4000 to 0x4300 can also be used for > common interrupt vector code. So, we can effectively use memory region > between 0x1800 to 0x4300 for common interrupt vector code. On Power9 the system-call-vectored instruction will use the region at 0x3000, so moving code into that space is not a good long term plan. I'll take your v2 and put it in next next week. cheers