Hi Jonathan, Thanks for the reply. But still there are some doubts, like while building bzImage, kbuild procedure does objcopy of /usr/src/linux/vmlinux to create /usr/src/linux/arch/i386/boot/compressed/vmlinux.bin ... [Here I have taken $temppiggy=vmlinux.bin]. It compresses it and then this it links it to create piggy.o :- $ gzip -9 < vmlinux.bin > vmlinux.bin.gz $ ld -r vmlinux.bin.gz -o piggy.o Now at this point I think that piggy.o is in ELF format, but there are only .data, .shstrtab, .symtab and .shmstr sections present in this. $ readelf -S piggy.o There are 5 section headers, starting at offset 0x10eb38: Section Headers: [Nr] Name Type Addr Off Size ES Flg Lk Inf Al [ 0] NULL 00000000 000000 000000 00 0 0 0 [ 1] .data PROGBITS 00000000 000034 10eae3 00 WA 0 0 1 [ 2] .shstrtab STRTAB 00000000 10eb17 000021 00 0 0 1 [ 3] .symtab SYMTAB 00000000 10ec00 0000b0 10 4 5 4 [ 4] .strtab STRTAB 00000000 10ecb0 000085 00 0 0 1 Key to Flags: W (write), A (alloc), X (execute), M (merge), S (strings) I (info), L (link order), G (group), x (unknown) O (extra OS processing required) o (OS specific), p (processor specific) After this kbuild compiles head.S and misc.c and gets head.o and misc.o. Now, it links these 2 with piggy.o to get bvmlinux... My question is if we don't have .text section at this point for vmlinux [which we had at starting], then how does kernel know about which section is data and which is text... ?? TIA. Sumit Sharma. On Sat, 18 Sep 2004 Jon Masters wrote : >On 14 Sep 2004 12:48:42 -0000, matrix reloaded ><matrix_reloaded18@xxxxxxxxxxxxxx> > > > I am going through the kbuild process and found that at some stage after > > building vmlinux inside "/usr/src/linux/" somewhere kbuild uses objcopy to build > > bzImage inside /usr/src/linux/arch/i386/boot/compressed :- > > > Could anyone please explain why we are creating binary image of vmlinux here. > >We want a compressed kernel image that can uncompress itself and >appear as an x86 boot sector image rather than a standard ELF >executable file. > >Linux build procedures tend to rely upon objcopy a lot to remove >unwanted sections from output files or to bodge in particular sections >- for example the dummy stage adds several utilities to an image file >which also contains the kernel image. I use objcopy on embedded boards >to bundle a kernel image with bootloader and firmware in to one >loadable image. > >Cheers, > >Jon.
