[Clfs-dev] CLFS Embedded

[Neil Bradley]

Hi Andrew,

Iptables is very useful, as is libnl & hostapd if you want your device to be a hotspot. This is functionality I will need in the near future, but I was okay with not installing these at the current stage. There might be a good chance the enthusiast would appreciate this kind of configurability. Is there a suitable alternative, or perhaps this belongs in CBLFS instead?

Regarding e2fsprogs & dropbear...
I haven't tested with the latest dropbear, I'll rebuild later this week, but the situation was thus:

I'm targeting ARM1176; the version of dropbear previously used had configuration scripts (config.guess and config.sub) dating from 2007, which don't support the architecture. I updated those scripts  to the latest versions from Savannah. I haven't tested with the new dropbear.

I'm not quite sure where the issue with e2fsprogs came in; I surmised that it was due to the target arch not having an unsigned 64-bit type. It's only used in 1 small section of ext2fs/unix_io.c that is for very large filesystems - but rather than just commenting it out, I rewrote it using an unsigned 32-bit type that the ARM supported.

lib/ext2fs/unix_io.c: 933 - 940:

#ifdef BLKDISCARD
        __uint64_t range[2];

        range[0] = (__uint64_t)(block) * channel->block_size;
        range[1] = (__uint64_t)(count) * channel->block_size;

        ret = ioctl(data->dev, BLKDISCARD, &range);
#else

Cheers, Neil


> Date: Thu, 17 Oct 2013 15:05:31 -0400
> From: andrew at bradfordembedded.com
> To: neil.bradley at windowslive.com
> Subject: Re: CLFS Embedded
> 

 ...
 
> 
> > 11.5.1 - iptables had same issue with config.? - but was in build-aux directory.
> > --enable-libipq misspelt?
> > Conflicts between net/if.h & linux/if.h : 
> > change net/if.h to linux/if.h in ipt_headers.... .h & include/xtable.h
> > also netinet/in.h to linux/in.h
> > also had u_int instead of uint
> > These issues in both 1.4.19.1 & 1.4.20.0
> 
> Yeah, programs that include both libc and linux headers are annoying
> with musl as it separates itself from Linux quite well (it's one of the
> only libc that doesn't rely on linux headers in order to build).
> Busybox does something similar, thus the somewhat ugly sed lines there
> to remove building of ifplugd.
> 
> I'd like to fix ifplugd, as it's a nice feature.
> 
> Is it worth keeping iptables for a normal embedded build?  I think this
> might again be more WRT leaking into the book for people wanting to
> build a router, which isn't really the direction I want to take the book.
> 
> > e2fsprogs: 
> > ioctl had an issue with 2x 64-bit block sizes in BLKDISCARD
> > changed to using 4x 32-bit in ext2fs/io_ctl.c
> 
> OK.  I'll open a trac ticket for this.
> 
> Can you write up the e2fsprogs and dropbear issues in a slightly cleaner
> way?  I'm not fully sure I fully grok them (and I've not yet build
> tested them recently).
> 
> Thanks!
> -Andrew



 		 	   		  
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.clfs.org/pipermail/clfs-dev-clfs.org/attachments/20131020/c66660e1/attachment.html>