[Top] [All Lists]

Re: [PATCH v2 1/2] MIPS: fix/enrich 34K APRP (APSP) functionalities

To: "Maciej W. Rozycki" <>
Subject: Re: [PATCH v2 1/2] MIPS: fix/enrich 34K APRP (APSP) functionalities
From: Deng-Cheng Zhu <>
Date: Tue, 22 May 2012 15:05:54 +0800
Cc: John Crispin <>, <>, <>
In-reply-to: <>
References: <> <> <> <> <> <> <> <>
User-agent: Mozilla/5.0 (X11; U; Linux i686; en-US; rv: Gecko/20120216 Lightning/1.0b2 Thunderbird/3.1.19
On 05/22/2012 07:17 AM, Maciej W. Rozycki wrote:
On Mon, 21 May 2012, Deng-Cheng Zhu wrote:

   What's so Malta-specific in the VPE loader anyway?  It's a CPU feature,
not a board-specific one.

Well, first off, for VPE loader itself, when it comes to CPS we have
vpe_run() that derives from amon_cpu_start() in arch/mips/mti-malta/malta-
amon.c. There is no implementation of amon_cpu_start() on other platforms.

  Hmm, there's nothing platform-specific there, the file is pretty generic,
it could be moved to arch/mips/kernel/ or thereabouts.  That applies to
<asm/mips-boards/launch.h>  too, before you ask

Yeah, agree with you. I didn't do it simply because I'm not sure :)

(you may want to use alloc_bootmem or suchlike instead of hardcoding the
trampoline page, though it's probably pretty safe to assume the end of
the exception handler page is available everywhere).

I'm not quite clear about this. Do you mean to bypass the arbitrary monitor
in vpe_run() (in other words, to directly bring up the vpe in vpe_run())?
Why do we need to worry about writing to the cpulaunch data?

Secondly, I suppose VPE loader works uniquely for APRP, and part of APRP
(such as IRQ related stuff) depends on platform code. So it makes sense
(IMO) to impose the dependency of APRP on the root (VPE loader).

  Hmm, does it really?  It sounds wrong to me, it shouldn't use any
hardware interrupts, and software interrupts again are available
everywhere, at least on the MT processors now in existence.

  There's nothing platform-specific referred to from arch/mips/kernel/vpe.c
AFAICT (and I trust in Beth having got this piece right).  I reckon it
used to work with CONFIG_MIPS_SIM too, though I could imagine the
configuration got neglected a bit as it is somewhat unusual.

Oh, When I said IRQ related stuff I meant the interrupt specific changes in
rtlx.c (not vpe.c) which correspond to those in malta-int.c. They are
there to resolve some issues (Please refer to the code changes and added
comments in these 2 files in PATCH #1 and #2.).

Thanks for your review,


<Prev in Thread] Current Thread [Next in Thread>