[Patch, avr] Fix PR ld/20545
Senthil Kumar Selvaraj
senthil_kumar.selvaraj@atmel.com
Fri Sep 2 06:33:00 GMT 2016
Hi,
The below patch fixes a couple of relaxation bugs in the linker for
the avr backend.
Prior to the patch, addends for relocs were being adjusted even if
they went beyond an alignment boundary. This is wrong - to
preserve alignment constraints, the relaxation logic adds as many padding
bytes at the alignment boundary as was deleted, so addends beyond the
boundary should not be adjusted. avr-prop-7.s reproduces this
scenario.
Also, prior to this patch, the relaxation logic assumed that the addr
parameter pointed to the middle of the instruction to be deleted, and
that addr - count would therefore be the shrinked instruction's
address. This is true when actually shrinking instructions.
The alignment constraints handling logic also invokes the same logic
though, with addr as the starting offset of padding bytes and
with count as the number of bytes to be deleted. Calculating the
shrinked insn's address as addr - count is obviously wrong in this
case - that offset would point to count bytes before the last
non-padded byte. avr-prop-8.s reproduces this scenario.
To fix scenario 1, the patch adds an additional check to ensure reloc addends
aren't adjusted if they cross a shrink boundary. The shrink boundary
is either the section size or an alignment boundary. Addends pointing
at an alignment boundary don't need to be adjusted, as padding would
occur and keep the boundary the same. Addends pointing at section size
need to be adjusted though, as no padding occurs and the section size
itself would get decremented. The patch records whether padding
occured (did_pad) and uses that to detect and handle this condition.
To fix scenario 2, the patch adds an additional parameter
(delete_shrinks_insn) to elf32_avr_relax_delete_bytes to distinguish
instruction bytes deletion from padding bytes deletion. It then uses that to
correctly set shrinked_insn_address.
Reg testing didn't show any regressions. Ok for trunk and backport to
2.27?
Regards
Senthil
bfd/ChangeLog:
2016-09-02 Senthil Kumar Selvaraj <senthil_kumar.selvaraj@atmel.com>
PR ld/20545
* elf32-avr.c (elf32_avr_relax_delete_bytes): Add parameter
delete_shrinks_insn. Modify computation of shrinked_insn_address.
Compute shrink_boundary and adjust addend only if
addend_within_shrink_boundary.
(elf32_avr_relax_section): Modify calls to
elf32_avr_relax_delete_bytes to pass extra parameter.
ld/ChangeLog:
2016-09-02 Senthil Kumar Selvaraj <senthil_kumar.selvaraj@atmel.com>
PR ld/20545
* testsuite/ld-avr/avr-prop-7.d: New test.
* testsuite/ld-avr/avr-prop-7.s: New test.
* testsuite/ld-avr/avr-prop-8.d: New test.
* testsuite/ld-avr/avr-prop-8.s: New test.
diff --git bfd/elf32-avr.c bfd/elf32-avr.c
index a0a5c69..eea76a4 100644
--- bfd/elf32-avr.c
+++ bfd/elf32-avr.c
@@ -1808,13 +1808,17 @@ elf32_avr_adjust_diff_reloc_value (bfd *abfd,
/* Delete some bytes from a section while changing the size of an instruction.
The parameter "addr" denotes the section-relative offset pointing just
behind the shrinked instruction. "addr+count" point at the first
- byte just behind the original unshrinked instruction. */
+ byte just behind the original unshrinked instruction. If delete_shrinks_insn
+ is FALSE, we are deleting redundant padding bytes from relax_info prop
+ record handling. In that case, addr is section-relative offset of start
+ of padding, and count is the number of padding bytes to delete. */
static bfd_boolean
elf32_avr_relax_delete_bytes (bfd *abfd,
asection *sec,
bfd_vma addr,
- int count)
+ int count,
+ bfd_boolean delete_shrinks_insn)
{
Elf_Internal_Shdr *symtab_hdr;
unsigned int sec_shndx;
@@ -1829,6 +1833,7 @@ elf32_avr_relax_delete_bytes (bfd *abfd,
struct avr_relax_info *relax_info;
struct avr_property_record *prop_record = NULL;
bfd_boolean did_shrink = FALSE;
+ bfd_boolean did_pad = FALSE;
symtab_hdr = &elf_tdata (abfd)->symtab_hdr;
sec_shndx = _bfd_elf_section_from_bfd_section (abfd, sec);
@@ -1909,6 +1914,7 @@ elf32_avr_relax_delete_bytes (bfd *abfd,
to remember we didn't delete anything i.e. don't set did_shrink,
so that we don't corrupt reloc offsets or symbol values.*/
memset (contents + toaddr - count, fill, count);
+ did_pad = TRUE;
/* Adjust the TOADDR to avoid moving symbols located at the address
of the property record, which has not moved. */
@@ -1965,7 +1971,9 @@ elf32_avr_relax_delete_bytes (bfd *abfd,
continue;
shrinked_insn_address = (sec->output_section->vma
- + sec->output_offset + addr - count);
+ + sec->output_offset + addr);
+ if (delete_shrinks_insn)
+ shrinked_insn_address -= count;
irel = elf_section_data (isec)->relocs;
/* PR 12161: Read in the relocs for this section if necessary. */
@@ -2002,6 +2010,13 @@ elf32_avr_relax_delete_bytes (bfd *abfd,
a symbol or section associated with it. */
if (sym_sec == sec)
{
+ /* If there is an alignment boundary, we only need to
+ adjust addends that end up below the boundary. */
+ bfd_vma shrink_boundary = (reloc_toaddr
+ + sec->output_section->vma
+ + sec->output_offset);
+ bfd_boolean addend_within_shrink_boundary = FALSE;
+
symval += sym_sec->output_section->vma
+ sym_sec->output_offset;
@@ -2015,8 +2030,17 @@ elf32_avr_relax_delete_bytes (bfd *abfd,
(unsigned int) (symval + irel->r_addend),
(unsigned int) shrinked_insn_address);
+ /* If we padded bytes, then the boundary didn't change,
+ so there's no need to adjust addends pointing at the boundary.
+ If we didn't pad, then we actually shrank the boundary, so
+ addends pointing at the boundary need to be adjusted too. */
+ addend_within_shrink_boundary = did_pad
+ ? ((symval + irel->r_addend) < shrink_boundary)
+ : ((symval + irel->r_addend) <= shrink_boundary);
+
if (symval <= shrinked_insn_address
- && (symval + irel->r_addend) > shrinked_insn_address)
+ && (symval + irel->r_addend) > shrinked_insn_address
+ && addend_within_shrink_boundary)
{
if (elf32_avr_is_diff_reloc (irel))
{
@@ -2648,7 +2672,8 @@ elf32_avr_relax_section (bfd *abfd,
{
/* Delete two bytes of data. */
if (!elf32_avr_relax_delete_bytes (abfd, sec,
- irel->r_offset + 2, 2))
+ irel->r_offset + 2, 2,
+ TRUE))
goto error_return;
/* That will change things, so, we should relax again.
@@ -2972,7 +2997,8 @@ elf32_avr_relax_section (bfd *abfd,
/* Delete two bytes of data. */
if (!elf32_avr_relax_delete_bytes (abfd, sec,
- irel->r_offset + insn_size, 2))
+ irel->r_offset + insn_size, 2,
+ TRUE))
goto error_return;
/* That will change things, so, we should relax
@@ -3040,7 +3066,7 @@ elf32_avr_relax_section (bfd *abfd,
record->offset -= count;
elf32_avr_relax_delete_bytes (abfd, sec,
addr - count,
- count);
+ count, FALSE);
*again = TRUE;
}
}
diff --git ld/testsuite/ld-avr/avr-prop-7.d ld/testsuite/ld-avr/avr-prop-7.d
new file mode 100644
index 0000000..9f2cea9
--- /dev/null
+++ ld/testsuite/ld-avr/avr-prop-7.d
@@ -0,0 +1,15 @@
+#name: AVR .avr.prop, AVR_7_PCREL after align
+#as: -mavrxmega2 -mlink-relax
+#ld: -mavrxmega2 --relax
+#source: avr-prop-7.s
+#objdump: -S
+#target: avr-*-*
+
+#...
+00000000 <__ctors_end>:
+ 0: 04 d0 rcall .+8 ; 0xa <foo>
+ 2: 00 00 nop
+ 4: 00 00 nop
+ 6: 86 e0 ldi r24, 0x06 ; 6
+ 8: f0 f7 brcc .-4 ; 0x6 <__ctors_end\+0x6>
+#...
diff --git ld/testsuite/ld-avr/avr-prop-7.s ld/testsuite/ld-avr/avr-prop-7.s
new file mode 100644
index 0000000..38276ba
--- /dev/null
+++ ld/testsuite/ld-avr/avr-prop-7.s
@@ -0,0 +1,8 @@
+ call foo
+ nop
+ .p2align 1
+ nop
+.L618:
+ ldi r24,lo8(6)
+ brsh .L618
+foo: nop
diff --git ld/testsuite/ld-avr/avr-prop-8.d ld/testsuite/ld-avr/avr-prop-8.d
new file mode 100644
index 0000000..2905f98
--- /dev/null
+++ ld/testsuite/ld-avr/avr-prop-8.d
@@ -0,0 +1,13 @@
+#name: AVR .avr.prop, AVR_7_PCREL just before align
+#as: -mavrxmega2 -mlink-relax
+#ld: -mavrxmega2 --relax
+#source: avr-prop-8.s
+#objdump: -S
+#target: avr-*-*
+
+#...
+00000000 <__ctors_end>:
+ 0: ff cf rjmp .-2 ; 0x0 <__ctors_end>
+ 2: fe df rcall .-4 ; 0x0 <__ctors_end>
+ 4: f8 f7 brcc .-2 ; 0x4 <__ctors_end\+0x4>
+#...
diff --git ld/testsuite/ld-avr/avr-prop-8.s ld/testsuite/ld-avr/avr-prop-8.s
new file mode 100644
index 0000000..34554f2
--- /dev/null
+++ ld/testsuite/ld-avr/avr-prop-8.s
@@ -0,0 +1,7 @@
+foo:
+ jmp foo
+ call foo
+.L1:
+ brsh .L1
+.p2align 1
+ nop
More information about the Binutils
mailing list