Re: BLCR and kernel 2.6.32

From: Paul H. Hargrove (PHHargrove_at_lbl_dot_gov)
Date: Sun Dec 20 2009 - 13:48:13 PST

  • Next message: Alan Woodland: "swp/swpne ARMv7"
    Alan Woodland wrote:
    > 2009/12/19 Alan Woodland <alan.woodland_at_gmail_dot_com>:
    >   
    >> 2009/12/18 Paul H. Hargrove <PHHargrove_at_lbl_dot_gov>:
    >>     
    >>> Paul H. Hargrove wrote:
    >>>       
    >>>> Alan Woodland wrote:
    >>>>         
    >>>>> 2009/12/18 Yuri D'Elia <yuri.delia_at_eurac_dot_edu>:
    >>>>>
    >>>>>           
    >>>>>> Just tried to see the dkms boot script in action...
    >>>>>>
    >>>>>> checking kernel symbol table for exported register_chrdev... not found
    >>>>>> ======================================================================
    >>>>>> Please review the following configuration information:
    >>>>>>  Kernel source directory = /lib/modules/2.6.32-trunk-amd64/build
    >>>>>>  Kernel build directory = /lib/modules/2.6.32-trunk-amd64/build
    >>>>>>  Kernel symbol table = /boot/System.map-2.6.32-trunk-amd64
    >>>>>>  Kernel version probed from kernel build = 2.6.32-trunk-amd64
    >>>>>>  Kernel running currently = 2.6.32-trunk-amd64
    >>>>>> ======================================================================
    >>>>>> configure: error: Unable to use kernel 2.6.32-trunk-amd64 - kernel does
    >>>>>> not export register_chrdev
    >>>>>>
    >>>>>> whoops :|
    >>>>>>
    >>>>>>             
    >>>>> Looks like things changed a bit in 2.6.32 again. I'm not going to have
    >>>>> time to look at it properly until the new year, although possibly
    >>>>> someone upstream has already encountered this one?
    >>>>>
    >>>>> Alan
    >>>>>
    >>>>>
    >>>>>           
    >>>> Regretfully, I also don't expect to look at this much until January.
    >>>> I'll try look today to see if it is something quick, but doubt that it
    >>>> will be.
    >>>>
    >>>> -Paul
    >>>>
    >>>>         
    >>> Good news.  This DID turn out to be a small/easy change.  With the attached
    >>> patch (in addition to the patch for 2.6.31 support) I can configure, build
    >>> and pass "make check" with a 2.6.32 kernel (from kernel.org) on a PPC64
    >>> machine.
    >>>       
    >> Thanks, I'll try and push an update to Debian sometime today with this
    >> patch applied.
    >>
    >>     
    > Seems to be a bit of a hiccup with the patch, build now fails with:
    >
    > make -C /lib/modules/2.6.32-trunk-amd64/build
    > SUBDIRS=/var/lib/dkms/blcr/0.8.2/build/blcr_imports/kbuild modules
    > CC="gcc"
    > make[1]: Entering directory `/usr/src/linux-headers-2.6.32-trunk-amd64'
    >   CC [M]  /var/lib/dkms/blcr/0.8.2/build/blcr_imports/kbuild/imports.o
    > {standard input}: Assembler messages:
    > {standard input}:5: Error: undefined symbol
    > `CR_EXPORTED_KCODE_register_chrdev' in operation
    > {standard input}:9: Error: undefined symbol
    > `CR_EXPORTED_KCODE_register_chrdev' in operation
    > {standard input}:13: Error: undefined symbol
    > `CR_EXPORTED_KCODE_register_chrdev' in operation
    > {standard input}:17: Error: undefined symbol
    > `CR_EXPORTED_KCODE_register_chrdev' in operation
    > {standard input}:21: Error: undefined symbol
    > `CR_EXPORTED_KCODE_register_chrdev' in operation
    > {standard input}:25: Error: undefined symbol
    > `CR_EXPORTED_KCODE_register_chrdev' in operation
    > {standard input}:29: Error: undefined symbol
    > `CR_EXPORTED_KCODE_register_chrdev' in operation
    > {standard input}:33: Error: undefined symbol
    > `CR_EXPORTED_KCODE_register_chrdev' in operation
    > {standard input}:37: Error: undefined symbol
    > `CR_EXPORTED_KCODE_register_chrdev' in operation
    > {standard input}:41: Error: undefined symbol
    > `CR_EXPORTED_KCODE_register_chrdev' in operation
    > {standard input}:45: Error: undefined symbol
    > `CR_EXPORTED_KCODE_register_chrdev' in operation
    > {standard input}:49: Error: undefined symbol
    > `CR_EXPORTED_KCODE_register_chrdev' in operation
    >
    > (repeats quite a bit more)
    >
    > If I've read it right I think include/blcr_ksyms.h needs to be patched
    > too, which wasn't touched in the patch you posted. I've attached my
    > quick hack that now builds and can be inserted into the kernel ok. I
    > will hold off on uploads for confirmation that it's sane and tests
    > results to finish running.
    >
    > Alan
    >   
    
    Alan is correct (as is his patch) about blcr_ksyms.h.
    I didn't notice the problem because I tested on PPC64 w/o 
    CONFIG_RELOCATABLE.
    
    Thanks,
    -Paul
    
    -- 
    Paul H. Hargrove                          PHHargrove_at_lbl_dot_gov
    Future Technologies Group                 Tel: +1-510-495-2352
    HPC Research Department                   Fax: +1-510-486-6900
    Lawrence Berkeley National Laboratory     
    

  • Next message: Alan Woodland: "swp/swpne ARMv7"