[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <535D45CE.7020309@pennware.com>
Date: Sun, 27 Apr 2014 13:00:46 -0500
From: Richard Pennington <rich@...nware.com>
To: musl@...ts.openwall.com
CC: Rich Felker <dalias@...c.org>
Subject: Re: Bare metal progress and an oddity.
On 04/27/2014 01:00 PM, Rich Felker wrote:
> On Sun, Apr 27, 2014 at 12:37:13PM -0500, Richard Pennington wrote:
>> As part of my quixotic quest to run musl on ARM bare metal I've
>> started to handle interrupts and an ARM timer to implement the
>> clock_* functions to prepare for pre-emptive scheduling of
>> pthread_create'd threads. The current sources are here:
>> http://ellcc.org/viewvc/svn/ellcc/trunk/baremetal/arm/
>>
>> An oddity: If I use printf() to print out values in an interrupt
>> handler it sort of works. The format string prints perfectly, but it
>> seems that the va_arg values are getting messed up. The stack is
>> large enough, fp and sp are sane. Am I missing something in
>> pop_arg() that is causing this? Come to think of it, do I have to
>> insure a certain stack alignment, for example? Most of the messed up
>> values are "long long".
>>
>> Any thoughts appreciated.
> Yes, ARM EABI requires the stack to be aligned to at least an 8-byte
> boundary (maybe 16 but I think 8 is all that's required). If your
> stack is only aligned to 4 bytes, this will mess things up badly.
>
> Rich
Rich,
You're absolutely right. As I was writing my initial email I thought of
the stack alignment. It turns out that my idle thread pushed 12 bytes on
the stack making it unaligned. The idle thread is a leaf function (of
course) so I guess clang didn't feel the need to keep the stack aligned.
After fixing the alignment, interrupt printf() works fine.
-Rich
Powered by blists - more mailing lists
Confused about mailing lists and their use?
Read about mailing lists on Wikipedia
and check out these
guidelines on proper formatting of your messages.
