Openwall GNU/*/Linux - a small security-enhanced Linux distro for servers
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date: Thu, 24 Jan 2013 14:19:22 -0500
From: Rich Felker <dalias@...ifal.cx>
To: musl@...ts.openwall.com
Subject: Re: unknown type name 'loff_t'

On Thu, Jan 24, 2013 at 07:57:28PM +0100, Alexander Stadler wrote:
> Am 23.01.2013 18:37, schrieb Rich Felker:
> > On Wed, Jan 23, 2013 at 06:33:42PM +0100, Alexander Stadler wrote:
> >> Thanks for your fast responses.
> >>
> >> The other error I'm currently running into is
> >>
> >> make[4]: Entering directory `/space/build-trunk/trunk/build_dir/target-mips_r2_musl-0.9.8/uboot-envtools-2012.04.01'
> >> mips-openwrt-linux-musl-gcc -Wall -Os -pipe -mips32r2 -mtune=mips32r2 -fno-caller-saves -mno-branch-likely -fhonour-copts -Wno-error=unused-but-set-variable -msoft-float  -I/space/build-trunk/trunk/staging_dir/target-mips_r2_musl-0.9.8/usr/include -I/space/build-trunk/trunk/staging_dir/target-mips_r2_musl-0.9.8/include -I/space/build-trunk/trunk/staging_dir/toolchain-mips_r2_gcc-4.6-linaro_musl-0.9.8/usr/include -I/space/build-trunk/trunk/staging_dir/toolchain-mips_r2_gcc-4.6-linaro_musl-0.9.8/include  crc32.c fw_env.c fw_env_main.c -o fw_printenv
> >> fw_env.c:643:55: error: unknown type name 'loff_t'
> >> fw_env.c: In function 'flash_read_buf':
> >> fw_env.c:680:2: error: unknown type name 'loff_t'
> >> fw_env.c:713:3: warning: implicit declaration of function 'flash_bad_block' [-Wimplicit-function-declaration]
> >> fw_env.c: In function 'flash_write_buf':
> >> fw_env.c:777:2: error: unknown type name 'loff_t'
> >> make[4]: *** [fw_printenv] Error 1
> >> make[4]: Leaving directory `/space/build-trunk/trunk/build_dir/target-mips_r2_musl-0.9.8/uboot-envtools-2012.04.01'
> > 
> > Where is loff_t being used? It's probably the wrong type to use. The
> > only APIs loff_t is to be used with are under _GNU_SOURCE, so if you
> > don't need -D_GNU_SOURCE to get the functions, it's doubtful that
> > loff_t should be used. These functions are in fcntl.h, and it will
> > #define loff_t off_t when _GNU_SOURCE is defined, so either
> > -D_GNU_SOURCE or -Dloff_t=off_t would be a workaround, but I think the
> > issue should be investigated to determine if there's code that needs
> > to be fixed or if there are other circumstances under which loff_t
> > should be exposed by musl.
> > 
> > Rich
> 
> It's used in
> http://git.denx.de/cgi-bin/gitweb.cgi?p=u-boot.git;a=blob;f=tools/env/fw_env.c;h=37b60b80a7624566d9ecefdd04c0658134bd654d;hb=HEAD
> where (you were absolutely right) a #include <fcntl.h> exists.
> And -D_GNU_SOURCE helped. Thank you!

It looks like some ioctl interfaces also request loff_t in their
official usage, so musl should do something to ensure it's exposed for
ioctl. I'm not sure however what the best way is. Simply adding

#define loff_t off_t

to sys/ioctl.h is not sufficient because off_t might not be defined,
so another header would need to be included to get off_t. Note that
sys/ioctl.h is not under the domain of any modern standard, so it
would not be a horrible offense to make it include extra junk if we
need to..

> Happy about that, unfortunately there will come up much more things
> I think, as seconds later the next undeclared error popped up..

OK, I'll take a look.

> (when compiling mtd
> mips-openwrt-linux-musl-gcc -Os -pipe -mips32r2 -mtune=mips32r2 -fno-caller-saves -mno-branch-likely -fhonour-copts -Wno-error=unused-but-set-variable -msoft-float -Dtarget_ar71xx=1 -Wall -DFIS_SUPPORT=1  -I/space/build-trunk/trunk/staging_dir/target-mips_r2_musl-0.9.8/usr/include -I/space/build-trunk/trunk/staging_dir/target-mips_r2_musl-0.9.8/include -I/space/build-trunk/trunk/staging_dir/toolchain-mips_r2_gcc-4.6-linaro_musl-0.9.8/usr/include -I/space/build-trunk/trunk/staging_dir/toolchain-mips_r2_gcc-4.6-linaro_musl-0.9.8/include  -Wall   -c -o fis.o fis.c
> fis.c: In function 'fis_open':
> fis.c:80:64: error: 'MAP_LOCKED' undeclared (first use in this function)
> fis.c:80:64: note: each undeclared identifier is reported only once for each function it appears in
> 
> So I wonder why uClibc? seems to include these things. But there are
> so many mman.h's and the one referenced from fis.c (sys/mman.h)
> seems not to have the define, but otheres have.. . How could a
> proper procedure look like to include these things, and on which
> side of the code this should go is what I currently don't understand
> at all. I just don't want to bother you if its bad behaviour of
> Uclibc which enabled this things to work. But than propably an
> generic workaround could be found.

MAP_LOCKED is a Linux extension to mmap that somehow never got added
to musl when the others were added. Its definition probably varies
per-arch. I'll see if I can get somebody to find what's needed to add
it correctly for all archs.

Rich

Powered by blists - more mailing lists

Your e-mail address:

Confused about mailing lists and their use? Read about mailing lists on Wikipedia and check out these guidelines on proper formatting of your messages.