Re: [leaf-user] ntpd segfaulting (copied to leaf-devel)
Hi Bob, Hi Erich; thanks for the profound work on this bug. I've contacted the maintainer of uClibc-ng and received some homework to track this issue. The proposed patch by by ddrown needs testing and if successful, might be accepted. kp Am Mittwoch, 3. Mai 2017, 22:04:53 schrieb Erich Titl: > Hi Bob > > (copying this to leaf-devel) > > Sorry, I was too fast, I did not look into the i386 code but the one in > x86_64 > > Am 03.05.2017 um 20:03 schrieb Robert K Coffman Jr. -Info From Data Corp.: > > Eric, > > > > IRC user ddrown confirmed the x64 patch I mentioned earlier seems to fix > > the issue on i386 as well. I'm not really sure what to do with that > > information. Is it possible to patch Leaf with it? > > > > https://gist.github.com/ddrown/15e943b8fe1da398320b0c0518c95554 > > I don't know. It looks like the RESTORE2 macro is extended with a nop > operator in this patch. This looks like assembly code. > > ... > > This is i386... > > #define RESTORE(name, syscall) RESTORE2(name, syscall) > > #ifdef __NR_rt_sigaction > /* The return code for realtime-signals. */ > # define RESTORE2(name, syscall) \ > __asm__ ( \ > ".text\n" \ > "__" #name ":\n"\ > " movl$" #syscall ", %eax\n" \ > " int $0x80\n"\ > ); > RESTORE(restore_rt, __NR_rt_sigreturn) > #endif > > #ifdef __NR_sigreturn > /* For the boring old signals. */ > # undef RESTORE2 > # define RESTORE2(name, syscall) \ > __asm__ ( \ > ".text\n" \ > "__" #name ":\n"\ > " popl%eax\n" \ > " movl$" #syscall ", %eax\n" \ > " int $0x80\n"\ > ); > RESTORE(restore, __NR_sigreturn) > #endif > > and this is x86_64 > > #define RESTORE(name, syscall) RESTORE2(name, syscall) > #define RESTORE2(name, syscall) \ > __asm__ ( \ > "nop\n" \ > ".text\n" \ > "__" #name ":\n"\ > " movq$" #syscall ", %rax\n" \ > " syscall\n" \ > ); > > #ifdef __NR_rt_sigaction > /* The return code for realtime-signals. */ > RESTORE(restore_rt, __NR_rt_sigreturn) > #endif > #ifdef __NR_sigreturn > RESTORE(restore, __NR_sigreturn) > #endif > > So in i386 the patch appears to be possible, if it is necessary I don't > know as it looks like the nop operator could be used to make the > assembly code align to instruction size. The nop operator is already in > the x86_64 code. > > I have not followed uClibc development. I am wondering what their > position is on this issue. > > cheers > > ET -- Check out the vibrant tech community on one of the world's most engaging tech sites, Slashdot.org! http://sdm.link/slashdot leaf-user mailing list: leaf-user@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/leaf-user Support Request -- http://leaf-project.org/
Re: [leaf-user] ntpd segfaulting (copied to leaf-devel)
Hi Bob (copying this to leaf-devel) Sorry, I was too fast, I did not look into the i386 code but the one in x86_64 Am 03.05.2017 um 20:03 schrieb Robert K Coffman Jr. -Info From Data Corp.: Eric, IRC user ddrown confirmed the x64 patch I mentioned earlier seems to fix the issue on i386 as well. I'm not really sure what to do with that information. Is it possible to patch Leaf with it? https://gist.github.com/ddrown/15e943b8fe1da398320b0c0518c95554 I don't know. It looks like the RESTORE2 macro is extended with a nop operator in this patch. This looks like assembly code. ... This is i386... #define RESTORE(name, syscall) RESTORE2(name, syscall) #ifdef __NR_rt_sigaction /* The return code for realtime-signals. */ # define RESTORE2(name, syscall) \ __asm__ ( \ ".text\n" \ "__" #name ":\n"\ " movl$" #syscall ", %eax\n" \ " int $0x80\n"\ ); RESTORE(restore_rt, __NR_rt_sigreturn) #endif #ifdef __NR_sigreturn /* For the boring old signals. */ # undef RESTORE2 # define RESTORE2(name, syscall) \ __asm__ ( \ ".text\n" \ "__" #name ":\n"\ " popl%eax\n" \ " movl$" #syscall ", %eax\n" \ " int $0x80\n"\ ); RESTORE(restore, __NR_sigreturn) #endif and this is x86_64 #define RESTORE(name, syscall) RESTORE2(name, syscall) #define RESTORE2(name, syscall) \ __asm__ ( \ "nop\n" \ ".text\n" \ "__" #name ":\n"\ " movq$" #syscall ", %rax\n" \ " syscall\n" \ ); #ifdef __NR_rt_sigaction /* The return code for realtime-signals. */ RESTORE(restore_rt, __NR_rt_sigreturn) #endif #ifdef __NR_sigreturn RESTORE(restore, __NR_sigreturn) #endif So in i386 the patch appears to be possible, if it is necessary I don't know as it looks like the nop operator could be used to make the assembly code align to instruction size. The nop operator is already in the x86_64 code. I have not followed uClibc development. I am wondering what their position is on this issue. cheers ET smime.p7s Description: S/MIME Cryptographic Signature -- Check out the vibrant tech community on one of the world's most engaging tech sites, Slashdot.org! http://sdm.link/slashdot leaf-user mailing list: leaf-user@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/leaf-user Support Request -- http://leaf-project.org/
Re: [leaf-user] ntpd segfaulting (copied to leaf-devel)
Hi Bob (copying this to leaf-devel) Am 03.05.2017 um 20:03 schrieb Robert K Coffman Jr. -Info From Data Corp.: Eric, IRC user ddrown confirmed the x64 patch I mentioned earlier seems to fix the issue on i386 as well. I'm not really sure what to do with that information. Is it possible to patch Leaf with it? https://gist.github.com/ddrown/15e943b8fe1da398320b0c0518c95554 I don't know. It looks like the RESTORE2 macro is extended with a nop operator in this patch. This looks like assembly code. The thing that puzzles me most is that the current code looks like the following. #define RESTORE(name, syscall) RESTORE2(name, syscall) #define RESTORE2(name, syscall) \ __asm__ ( \ "nop\n" \ ".text\n" \ "__" #name ":\n"\ " movq$" #syscall ", %rax\n" \ " syscall\n" \ ); So it looks like the patch has already been applied. Very definitely ddrown has not started from the same codebase, so I would be very reluctant to follow this path. cheers ET smime.p7s Description: S/MIME Cryptographic Signature -- Check out the vibrant tech community on one of the world's most engaging tech sites, Slashdot.org! http://sdm.link/slashdot leaf-user mailing list: leaf-user@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/leaf-user Support Request -- http://leaf-project.org/
