On Fri, Aug 16, 2013 at 10:32 PM, Arik Nemtsov <arik@xxxxxxxxxx> wrote: > On Fri, Aug 16, 2013 at 8:26 PM, Hauke Mehrtens <hauke@xxxxxxxxxx> wrote: >> This fixes a problem introduced in this commit: >> commit c871780b5afa182878884bf5ccd8df4817a2660f >> Author: Arik Nemtsov <arik@xxxxxxxxxx> >> Date: Wed Aug 14 10:48:05 2013 +0300 >> >> backports: rename some mem functions to not break custom kernels >> >> Signed-off-by: Hauke Mehrtens <hauke@xxxxxxxxxx> >> Cc: Arik Nemtsov <arik@xxxxxxxxxx> > > Thanks for the fix (again). Indeed it was stupid of me to make the > #ifndef redundant. > I'll test it on my setup next week. Ok. This doesn't work for me, and it's pretty obvious why that I look at the code. I'm using a custom 3.9 kernel that contains the patch adding "arch_phys_wc_add". If indeed "arch_phys_wc_add" is defined (as is my case), we don't enter the clause at all and the LINUX_BACKPORT() definition doesn't happen. But the code in backport-3.11.c does the following anyway: EXPORT_SYMBOL_GPL(arch_phys_wc_add); So we still get the exported symbol "arch_phys_wc_add", clashing with the kernel's. Maybe some new define should be added to the backports header file to control when backport-3.11.c exports the symbol? Something in io.h along the lines of: #ifndef arch_phys_wc_add #define BACKPORTS_DEFINE_arch_phys_wc_add ... and then condition the symbol export in the .c file on BACKPORTS_DEFINE_arch_phys_wc_add. Arik -- To unsubscribe from this list: send the line "unsubscribe backports" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
