Date: Tue, 21 May 2013 19:05:09 +0200 From: Zbyszek Bodek <zbb@semihalf.com> To: Daan Vreeken <Daan@vitsch.nl> Cc: freebsd-arm@freebsd.org, Alan Cox <alc@cs.rice.edu>, ray@freebsd.org Subject: Re: New pmap-v6.c features and improvements Message-ID: <519BA945.8090908@semihalf.com> In-Reply-To: <201305211659.36599.Daan@vitsch.nl> References: <519b6b1c.9060008@semihalf.com> <1369143378-3800692384.5b4e94954b@bliksem.vehosting.nl> <519B784B.309@semihalf.com> <201305211659.36599.Daan@vitsch.nl>
next in thread | previous in thread | raw e-mail | index | archive | help
On 21.05.2013 16:59, Daan Vreeken wrote: > Hi Zbyszek, > > On Tuesday 21 May 2013 15:36:11 Zbyszek Bodek wrote: >> On 21.05.2013 15:15, Daan Vreeken wrote: >>> Hi Zbyszek, >>> >>> On Tuesday 21 May 2013 14:39:56 Zbyszek Bodek wrote: >>>> Hello Everyone, >>>> >>>> I would like to introduce another pack of patches for pmap-v6.c and >>>> related, that we created as a part of Semihalf work on Superpages >>>> support. >>>> >>>> The patches include some major changes like: >>> .. >>>> And some minor clean-ups: >>>> - Get rid of the VERBOSE_INIT_ARM option >>> .. >>> Just a small request: Could you leave the VERBOSE_INIT_ARM printf()'s >>> intact? VERBOSE_INIT_ARM is typically one of the first things I enable >>> when I'm hunting down problems during board bring-up. >> >> To which files are you referring? >> >> VERBOSE_INIT_ARM is useless in pmap as pmap functions that are using >> VERBOSE_INIT_ARM are called before the console is initialized so no >> benefit from printfs... >> >> What is your opinion. > > > This is the console output of a test kernel that I've created a while back for > an Atmel at91sam9261 processor board (with VERBOSE_INIT_ARM enabled) : > > ## Starting application at 0x200000E0 ... > DAAN: soc: at91sam9261 > DAAN: pre devmap bootstrap > devmap: fff00000 -> ffffffff @ dff00000 > pmap_map_chunk: pa=0xfff00000 va=0xdff00000 size=0x100000 resid=0x100000 > prot=0x3 cache=0 > S > devmap: 00500000 -> 005fffff @ dfc00000 > pmap_map_chunk: pa=0x500000 va=0xdfc00000 size=0x100000 resid=0x100000 > prot=0x3 cache=0 > S > devmap: 10000000 -> 1fffffff @ e0000000 > pmap_map_chunk: pa=0x10000000 va=0xe0000000 size=0x10000000 resid=0x10000000 > prot=0x3 cache=0 > SSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSS > devmap: 30000000 -> 300fffff @ dfb00000 > pmap_map_chunk: pa=0x30000000 va=0xdfb00000 size=0x100000 resid=0x100000 > prot=0x3 cache=0 > S > devmap: 00600000 -> 006fffff @ dfa00000 > pmap_map_chunk: pa=0x600000 va=0xdfa00000 size=0x100000 resid=0x100000 > prot=0x3 cache=0 > S > DAAN: post devmap bootstrap > DAAN: memsize=67108864 > GDB: debug ports: uart > GDB: current port: uart > KDB: debugger backends: ddb gdb > KDB: current backend: ddb > Copyright (c) 1992-2012 The FreeBSD Project. > ... > > > > To be able to create one kernel for multiple boards, we've changed > at91_machdep.c to enable a minimalistic pagetable early on in initarm() so > the SoC type can be identified. (As a side effect of this, the console works > a bit earlier than normal..) > Depending on the type of SoC detected, a different pagetable will be passed to > pmap_devmap_bootstrap(). At this time, we have a working console. While > working on this code I've made good use of the VERBOSE_INIT_ARM option. > > > Regards, > Hello again Daan, OK, I understand now. This is not applicable to the new common initarm() but if you are using this option then indisputably it should be preserved. Thanks for pointing that out! VERBOSE_INIT_ARM removal is not a crucial change and we will exclude it from the patch :). Do you have any other comments? Best regards Zbyszek Bodek
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?519BA945.8090908>