Don't fudge p_vaddr when PHDR in segment

RX does horrible fudges to PT_LOAD p_vaddr, that affect the testsuite
and mean the target won't support dynamic objects.  The latter
probably doesn't matter too much since RX is an embedded target, but
it's easy to stop some of the fudges in order to reduce special cases
for RX in the testsuite.  The changes make sense in isolation too.

bfd/
	* elf32-rx.c (elf32_rx_modify_program_headers): Don't adjust
	segments that include the ELF file header or program headers.
ld/
	* testsuite/ld-elf/flags1.d: Run for RX.
	* testsuite/ld-scripts/phdrs.exp: Likewise.
	* testsuite/ld-scripts/pr14962.d: Likewise.
	* testsuite/ld-scripts/pr14962-2.d: Likewise.
This commit is contained in:
Alan Modra 2016-12-13 09:29:47 +10:30
parent ea0ffd5be2
commit 6392030005
7 changed files with 29 additions and 12 deletions

View File

@ -1,3 +1,8 @@
2016-12-13 Alan Modra <amodra@gmail.com>
* elf32-rx.c (elf32_rx_modify_program_headers): Don't adjust
segments that include the ELF file header or program headers.
2016-12-08 Alan Modra <amodra@gmail.com>
PR ld/20932

View File

@ -3200,9 +3200,11 @@ rx_elf_object_p (bfd * abfd)
int i;
unsigned int u;
Elf_Internal_Phdr *phdr = elf_tdata (abfd)->phdr;
int nphdrs = elf_elfheader (abfd)->e_phnum;
Elf_Internal_Ehdr *ehdr = elf_elfheader (abfd);
int nphdrs = ehdr->e_phnum;
sec_ptr bsec;
static int saw_be = FALSE;
bfd_vma end_phdroff;
/* We never want to automatically choose the non-swapping big-endian
target. The user can only get that explicitly, such as with -I
@ -3227,6 +3229,17 @@ rx_elf_object_p (bfd * abfd)
corresponds (based on matching file offsets) and use its VMA
information to reconstruct the p_vaddr field we clobbered when we
wrote it out. */
/* If PT_LOAD headers include the ELF file header or program headers
then the PT_LOAD header does not start with some section contents.
Making adjustments based on the difference between sh_offset and
p_offset is nonsense in such cases. Exclude them. Note that
since standard linker scripts for RX do not use SIZEOF_HEADERS,
the linker won't normally create PT_LOAD segments covering the
headers so this is mainly for passing the ld testsuite.
FIXME. Why are we looking at non-PT_LOAD headers here? */
end_phdroff = ehdr->e_ehsize;
if (ehdr->e_phoff != 0)
end_phdroff = ehdr->e_phoff + nphdrs * ehdr->e_phentsize;
for (i=0; i<nphdrs; i++)
{
for (u=0; u<elf_tdata(abfd)->num_elf_sections; u++)
@ -3234,6 +3247,7 @@ rx_elf_object_p (bfd * abfd)
Elf_Internal_Shdr *sec = elf_tdata(abfd)->elf_sect_ptr[u];
if (phdr[i].p_filesz
&& phdr[i].p_offset >= end_phdroff
&& phdr[i].p_offset <= (bfd_vma) sec->sh_offset
&& sec->sh_size > 0
&& sec->sh_type != SHT_NOBITS

View File

@ -1,3 +1,10 @@
2016-12-13 Alan Modra <amodra@gmail.com>
* testsuite/ld-elf/flags1.d: Run for RX.
* testsuite/ld-scripts/phdrs.exp: Likewise.
* testsuite/ld-scripts/pr14962.d: Likewise.
* testsuite/ld-scripts/pr14962-2.d: Likewise.
2016-12-08 Alan Modra <amodra@gmail.com>
* configure: Regenerate.

View File

@ -2,7 +2,6 @@
#ld: -Tflags1.ld
#objcopy_linked_file: --set-section-flags .post_text_reserve=contents,alloc,load,readonly,code
#readelf: -S --wide
#notarget: rx-*-*
#...
Section Headers:

View File

@ -59,14 +59,6 @@ if ![ld_simple_link $ld tmpdir/phdrs $ldopt] {
set exec_output [prune_warnings $exec_output]
verbose -log $exec_output
# The RX port sets the p_paddr of loadable segments to 0 in order
# to be compatible with Renesas tools. When an RX executable is
# loaded into a BFD based tool the code tries to reconstruct the
# correct vaddr and paddr values. This is not always possible
# however and this test is one example of where the reconstruction
# fails.
setup_xfail rx-*-*
if [regexp $phdrs_regexp $exec_output] {
pass $testname
} else {

View File

@ -1,7 +1,7 @@
#ld: -T pr14962-2.t
#source: pr14962a.s
#nm: -n
#notarget: rx-*-* frv-*-*linux*
#notarget: frv-*-*linux*
#...
0+2000 [AT] _start

View File

@ -2,7 +2,7 @@
#source: pr14962a.s
#source: pr14962b.s
#nm: -n
#notarget: rx-*-* mmix-knuth-mmixware
#notarget: mmix-knuth-mmixware
# The reference to x would normally generate a cross-reference error
# but the linker script converts x to absolute to avoid the error.