Hi Robert, On Sat, 1 Aug 2009 12:28:48 +0530, Siddu <siddu.sjce@xxxxxxxxx> wrote: > On Fri, Jul 31, 2009 at 2:48 AM, Robert P. J. Day > <rpjday@xxxxxxxxxxxxxx>wrote: > >> >> as a followup to my earlier post, here's something i want to >> clarify. there's a difference in how much symbol-based debugging you >> can do based on whether you configured your running kernel with >> CONFIG_DEBUG_INFO or not, so i wanted to test what the differences > > > Can someone tell me where to mark CONFIG_DEBUG_INFO=y ? > > >> >> were between two kernels, one configured with, one without. >> >> after i built both kernels, you can see the predictable difference >> in size (ignore vmlinux.o, of course): >> >> $ ls -l vmlinux* >> -rwxrwxr-x. 1 rpjday rpjday 15649485 2009-07-30 16:51 vmlinux >> -rwxrwxr-x. 1 rpjday rpjday 116565773 2009-07-30 16:34 vmlinux.debug.info >> -rw-rw-r--. 1 rpjday rpjday 15891639 2009-07-30 16:51 vmlinux.o >> $ >> >> my basic question with the above is whether, apart from the >> additional debug info embedded in the kernel image, were the two >> kernels *exactly* the same -- that is, would they have the same >> sections, with the same sizes, at the same offsets, so i could bounce >> from one to the other with "gdb" to test the differences and not have >> to keep rebooting? >> >> as a quick test, i ran >> >> $ objdump --section-headers <kernel image name> >> >> on both, and the only difference was the additional debug sections at >> the end of the debug version: >> >> 2c2 >> < vmlinux: file format elf64-x86-64 >> --- >> > vmlinux.debug.info: file format elf64-x86-64 >> 89a90,107 >> > 42 .debug_aranges 0001b040 0000000000000000 0000000000000000 >> > 00cace00 >> 2**4 >> > CONTENTS, READONLY, DEBUGGING >> > 43 .debug_pubnames 00051010 0000000000000000 0000000000000000 >> 00cc7e40 2**0 >> > CONTENTS, READONLY, DEBUGGING >> > 44 .debug_info 04c22970 0000000000000000 0000000000000000 >> > 00d18e50 >> 2**0 >> > CONTENTS, READONLY, DEBUGGING >> > 45 .debug_abbrev 001e0d1e 0000000000000000 0000000000000000 >> > 0593b7c0 >> 2**0 >> > CONTENTS, READONLY, DEBUGGING >> > 46 .debug_line 0044f12b 0000000000000000 0000000000000000 >> > 05b1c4de >> 2**0 >> > CONTENTS, READONLY, DEBUGGING >> > 47 .debug_frame 0013ed00 0000000000000000 0000000000000000 >> > 05f6b610 >> 2**3 >> > CONTENTS, READONLY, DEBUGGING >> > 48 .debug_str 0018f17c 0000000000000000 0000000000000000 >> > 060aa310 >> 2**0 >> > CONTENTS, READONLY, DEBUGGING >> > 49 .debug_loc 0090b310 0000000000000000 0000000000000000 >> > 0623948c >> 2**0 >> > CONTENTS, READONLY, DEBUGGING >> > 50 .debug_ranges 001a5f10 0000000000000000 0000000000000000 >> > 06b447a0 >> 2**4 >> > CONTENTS, READONLY, DEBUGGING >> >> the fact that all of the prior sections were (apparently) identical >> suggests that i can, in fact, swap one kernel image with the other for >> "gdb" and everything should continue to work, with the exception of >> having symbols or not. is that a safe assumption? it *seems* to be, >> i'm just wondering if anyone knows for sure. >> >> rday >> >> p.s. the above suggests that, if those debug sections are the only >> difference, rather than build both versions of the kernel, i could >> build the DEBUG_INFO version, save that, then strip a copy of it of >> all of those debug sections to get the same effect, yes? >> >> -- >> >> ======================================================================== >> Robert P. J. Day Waterloo, Ontario, CANADA >> >> Linux Consulting, Training and Annoying Kernel Pedantry. >> >> Web page: http://crashcourse.ca >> Twitter: http://twitter.com/rpjday >> "Kernel Newbie Corner" column @ linux.com: http://cli.gs/WG6WYX >> ======================================================================== >> >> -- >> To unsubscribe from this list: send an email with >> "unsubscribe kernelnewbies" to ecartis@xxxxxxxxxxxx >> Please read the FAQ at http://kernelnewbies.org/FAQ >> >> >> the fact that all of the prior sections were (apparently) identical >> suggests that i can, in fact, swap one kernel image with the other for >> "gdb" and everything should continue to work, with the exception of >> having symbols or not. is that a safe assumption? it *seems* to be, >> i'm just wondering if anyone knows for sure. I haven't run gdb server yet with a full debug enabled kernel, but a stripped kernel runs fine on gbd. You just have to ensure you don't jump into the kernel code - even in dissambly-only, it causes an unrecoverable error for some reason I haven't figured out yet. I've run user apps from Insight and from Eclipse Galileo with CDT 6.0. Both debug perfectly fine, using SSH. The only difference being that Eclipse doesn't require you to first start gdbserver on the target (it remotely launches gdb from Eclipse as long as sshd is SFTP enabled - Dropbear isn't, but OpenSSH of course is). HTH B rgds Kris -- To unsubscribe from this list: send an email with "unsubscribe kernelnewbies" to ecartis@xxxxxxxxxxxx Please read the FAQ at http://kernelnewbies.org/FAQ
