On Mon 2019-02-18 17:22, Petr wrote:

On Mon 2019-02-18 13:29:11, chengjian (D) wrote:

...

Hi，Jiri

This patch should be merged into 4.4 stable,

which still use klp_write_module_reloc.

https://elixir.bootlin.com/linux/v4.4.174/source/arch/x86/kernel/livep atch.c

ZeFeng may have sent a stable(4.4-y) patch to the wrong mail-list(mainline).

ZeFeng or Chengjian, please, send the patch once again with stable@vger.kernel.org in CC and explanation that it is needed only for 4.4 and why.

This thread is already too long and messed to be proceed by stable people effectively.

Best Regards, Petr

On x86-64, for 32-bit PC-relacive branches, we can generate PLT32 relocation, instead of PC32 relocation. and R_X86_64_PLT32 can be treated the same as R_X86_64_PC32 since linux kernel doesn't use PLT.

commit b21ebf2fb4cd ("x86: Treat R_X86_64_PLT32 as R_X86_64_PC32") been fixed for the module loading, but not fixed for livepatch relocation, which will fail to load livepatch with the error message as follow: relocation failed for symbol <symbol name> at <symbol address>

This issue only effacted the kernel version from 4.0 to 4.6, becauce the function klp_write_module_reloc is introduced by: commit b700e7f03df5 ("livepatch: kernel: add support for live patching") and deleted by: commit 425595a7fc20 ("livepatch: reuse module loader code to write relocations")

Signed-off-by: chenzefeng chenzefeng2@huawei.com --- arch/x86/kernel/livepatch.c | 1 + 1 file changed, 1 insertion(+)

diff --git a/arch/x86/kernel/livepatch.c b/arch/x86/kernel/livepatch.c index d1d35cc..579f8f8 100644 --- a/arch/x86/kernel/livepatch.c +++ b/arch/x86/kernel/livepatch.c @@ -58,6 +58,7 @@ int klp_write_module_reloc(struct module *mod, unsigned long type, val = (s32)value; break; case R_X86_64_PC32: + case R_X86_64_PLT32: val = (u32)(value - loc); break; default: -- 1.8.5.6