From: Frank Barknecht Subject: Re: vmalloc_to_page and building alsa 0.9.2 on kernel 2.4.18 Date: Tue, 22 Apr 2003 16:56:33 +0200 Sender: alsa-devel-admin@lists.sourceforge.net Message-ID: <20030422145633.GA17186@fliwatut.scifi> References: <20030422141031.GA9845@fliwatut.scifi> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Return-path: Content-Disposition: inline In-Reply-To: Errors-To: alsa-devel-admin@lists.sourceforge.net List-Help: List-Post: List-Subscribe: , List-Unsubscribe: , List-Archive: To: Takashi Iwai Cc: alsa-devel@lists.sourceforge.net, 190052@bugs.debian.org List-Id: alsa-devel@alsa-project.org Hallo, Takashi Iwai hat gesagt: // Takashi Iwai wrote: > > I'm struggling with building alsa-0.9.2 on my trusted old kernel > > 2.4.18. Here's some info, cc'ed to the Debian-ALSA team. Regardless of > > what I do I always get unresolved symbols for snd-page-alloc: > > > > # depmod -ae > > depmod: *** Unresolved symbols in /lib/modules/2.4.18/alsa/snd-page-alloc.o > > depmod: vmalloc_to_page > > > > > > This is a warning I get for this while compiling the modules with > > gcc-3.2.3: > > > > sgbuf.c: In function `store_page_tables': > > sgbuf.c:70: warning: implicit declaration of function `vmalloc_to_page' > > sgbuf.c:70: warning: assignment makes pointer from integer without a cast > > > > I'm using the sources from the Debian packages, but as I read about a > > bug in the Makefiles, for which Clemens submitted a patch, I also > > tried building with the (a)core Makefiles from CVS, but without > > success. > > > > Then I searched the web, what this "vmalloc_to_page" is, and I found, > > that it should be defined in the Kernel-sources. But a "grep -rl > > vmalloc_to_page /usr/src/linux/" returns no hits! So 2.4.18 doesn't > > seem to have this function. > > > > This also is, what the configure script detects. From "config.log": > > > > configure:3735: checking for vmalloc_to_page > > configure:3761: gcc -c -g -O2 -I/usr/src/linux/include conftest.c >&5 > > configure: In function `main': > > configure:3753: `vmalloc_to_page' undeclared (first use in this function) > > configure:3753: (Each undeclared identifier is reported only once > > configure:3753: for each function it appears in.) > > > > > > Now alsa-driver/acore/sgbuf.c checks for a version number, before > > using vmalloc_to_page, for example like this: > > > > #if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 4, 0) > > page = vmalloc_to_page(vmaddr + (i << PAGE_SHIFT)); > > > > > > I don't know if that matters, though, it just raised my interest: > > Could this be a wrong check? > > > > And more importantly to me: How can I get alsa-0.9.2 to work on > > 2.4.18? > > a vmalloc_to_page() wrapper is defined in the alsa-driver tree for the > older kernels. please make sure that you run "make clean" before > building all. Well, I did that, and actually the debian-scripts force one to do it. I even removed the whole alsa-driver source tree. ;) It seems, that this wrapper somehow isn't correctly called. At least I get this unresolved symbols on my 2.4.18. > anyway, i would recommend you to try the cvs version. > > also, please run "modprobe -r snd-page-alloc" manually after calling > "/etc/init.d/alsasound stop". this module will be retained even after > alsa service is stopped. But it didn't even load ;( Anyway: following your advice I now built the cvs-modules, because I just cannot be without sound for more than two days ;) So somehow _my_ problem is solved, but I suspect that it will bite other users as well. Thanks for your help. ciao -- Frank Barknecht _ ______footils.org__ ------------------------------------------------------- This sf.net email is sponsored by:ThinkGeek Welcome to geek heaven. http://thinkgeek.com/sf