Skip to contentSkip to conversation

remove check for ELF offset in powerpc bootimage

PATCH In review Olaf Hering 2 participants 2 messages Jan 30, 2006 PowerPC

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 remove check for ELF offset in powerpc bootimage

Is this check really needed, are there PT_LOAD sections with offset zero (either zImage or vmlinux)? I see an offset which is always 64k.

Do not check for offset, it is always set.

 arch/powerpc/boot/main.c |    4 ++--
 1 files changed, 2 insertions(+), 2 deletions(-)
arch/powerpc/boot/main.c +2 -2
Index: linux-2.6.16-rc1-olh/arch/powerpc/boot/main.c
===================================================================
@@ -152,7 +152,7 @@ static int is_elf64(void *hdr)
 	elf64ph = (Elf64_Phdr *)((unsigned long)elf64 +
 				 (unsigned long)elf64->e_phoff);
 	for (i = 0; i < (unsigned int)elf64->e_phnum; i++, elf64ph++)
-		if (elf64ph->p_type == PT_LOAD && elf64ph->p_offset != 0)
+		if (elf64ph->p_type == PT_LOAD)
 			break;
 	if (i >= (unsigned int)elf64->e_phnum)
 		return 0;
@@ -193,7 +193,7 @@ static int is_elf32(void *hdr)
 	elf32 = (Elf32_Ehdr *)elfheader;
 	elf32ph = (Elf32_Phdr *) ((unsigned long)elf32 + elf32->e_phoff);
 	for (i = 0; i < elf32->e_phnum; i++, elf32ph++)
-		if (elf32ph->p_type == PT_LOAD && elf32ph->p_offset != 0)
+		if (elf32ph->p_type == PT_LOAD)
 			break;
 	if (i >= elf32->e_phnum)
 		return 0;
Signature · 2 lines
short story of a lazy sysadmin:
 alias appserv=wotan
#1 lore raw

Hi,

Quoted 2 lines · Olaf Hering

Is this check really needed, are there PT_LOAD sections with offset zero (either zImage or vmlinux)? I see an offset which is always 64k.

Sure, an offset of zero is completely legal (obviously, the start address would then be different from the load address of that segment).

Without context, I would say this loop is used to find some loadable segment that does not include the ELF header. I cannot think of any application for that, as it is certainly acceptable to merge LOAD segments, and the segment it is looking for could have been merged with the segment that includes the header.

If the current linker scripts specify that loading the header is unneeded, this means that all LOAD headers will have nonzero offsets. But I wouldn't count on that forever.[1]

Simon

[1] the Amiga bootloader, for example, assumes that all segments are to be loaded, so things broke badly on APUS when the ldscript did not remove the .note.gnu-stack section, which has a VMA of zero as it is never loaded anyway. So depending on things being a particular way is bad. :-)