[v2] bfd: Discard region regardless of warning flag
Checks
Commit Message
v1 -> v2:
Added testcase
Ok for trunk?
I have commit access for GDB. Do I still need someone to push this for
me or do I automatically have write access to the complete binutils
repository and just need the approval to push it myself?
---
The discard of the region should be performed regardless if the warning
for the discard is enabled or not.
Without this patch, ld would segfault in bfd_section_removed_from_list,
called in the if-statement right after this block, as the argument
isec->output_section can be NULL.
Signed-off-by: Torbjörn SVENSSON <torbjorn.svensson@foss.st.com>
Co-Authored-By: Yvan ROUX <yan.roux@foss.st.com>
---
bfd/elflink.c | 10 +++----
ld/testsuite/ld-arm/arm-elf.exp | 1 +
ld/testsuite/ld-arm/non-contiguous-arm7.d | 4 +++
ld/testsuite/ld-arm/non-contiguous-arm7.ld | 32 ++++++++++++++++++++++
ld/testsuite/ld-arm/non-contiguous-arm7.s | 16 +++++++++++
5 files changed, 58 insertions(+), 5 deletions(-)
create mode 100644 ld/testsuite/ld-arm/non-contiguous-arm7.d
create mode 100644 ld/testsuite/ld-arm/non-contiguous-arm7.ld
create mode 100644 ld/testsuite/ld-arm/non-contiguous-arm7.s
Comments
On Mon, Dec 19, 2022 at 07:34:51PM +0100, Torbjörn SVENSSON via Binutils wrote:
> v1 -> v2:
> Added testcase
>
> Ok for trunk?
This is OK except for a few minor things.
> I have commit access for GDB. Do I still need someone to push this for
> me or do I automatically have write access to the complete binutils
> repository and just need the approval to push it myself?
You should be able to push it.
> The discard of the region should be performed regardless if the warning
This should say "The discard of symbols in..." not "The discard of...".
The code here is handling symbols.
> for the discard is enabled or not.
> +++ b/ld/testsuite/ld-arm/non-contiguous-arm7.d
> @@ -0,0 +1,4 @@
> +#name: non-contiguous-arm7
> +#source: non-contiguous-arm7.s
> +#ld: --enable-non-contiguous-regions -T non-contiguous-arm7.ld --gc-sections
Please remove --gc-sections. The testcase doesn't need it.
On 2022-12-19 23:59, Alan Modra wrote:
> On Mon, Dec 19, 2022 at 07:34:51PM +0100, Torbjörn SVENSSON via Binutils wrote:
>> v1 -> v2:
>> Added testcase
>>
>> Ok for trunk?
>
> This is OK except for a few minor things.
>
>> I have commit access for GDB. Do I still need someone to push this for
>> me or do I automatically have write access to the complete binutils
>> repository and just need the approval to push it myself?
>
> You should be able to push it.
>
>> The discard of the region should be performed regardless if the warning
>
> This should say "The discard of symbols in..." not "The discard of...".
> The code here is handling symbols.
>
>> for the discard is enabled or not.
>
>> +++ b/ld/testsuite/ld-arm/non-contiguous-arm7.d
>> @@ -0,0 +1,4 @@
>> +#name: non-contiguous-arm7
>> +#source: non-contiguous-arm7.s
>> +#ld: --enable-non-contiguous-regions -T non-contiguous-arm7.ld --gc-sections
>
> Please remove --gc-sections. The testcase doesn't need it.
>
Pushed with the modifications mentioned above.
@@ -11154,12 +11154,12 @@ elf_link_input_bfd (struct elf_final_link_info *flinfo, bfd *input_bfd)
if (isym->st_shndx != SHN_UNDEF
&& isym->st_shndx < SHN_LORESERVE
&& isec->output_section == NULL
- && flinfo->info->non_contiguous_regions
- && flinfo->info->non_contiguous_regions_warnings)
+ && flinfo->info->non_contiguous_regions)
{
- _bfd_error_handler (_("warning: --enable-non-contiguous-regions "
- "discards section `%s' from '%s'\n"),
- isec->name, bfd_get_filename (isec->owner));
+ if (flinfo->info->non_contiguous_regions_warnings)
+ _bfd_error_handler (_("warning: --enable-non-contiguous-regions "
+ "discards section `%s' from '%s'\n"),
+ isec->name, bfd_get_filename (isec->owner));
continue;
}
@@ -1271,6 +1271,7 @@ run_dump_test "non-contiguous-arm3"
run_dump_test "non-contiguous-arm4"
run_dump_test "non-contiguous-arm5"
run_dump_test "non-contiguous-arm6"
+run_dump_test "non-contiguous-arm7"
if { !$is_nacl && [check_shared_lib_support] } {
run_dump_test "thumb-plt"
new file mode 100644
@@ -0,0 +1,4 @@
+#name: non-contiguous-arm7
+#source: non-contiguous-arm7.s
+#ld: --enable-non-contiguous-regions -T non-contiguous-arm7.ld --gc-sections
+#error: \A.*unresolvable R_ARM_ABS32 relocation against symbol .MY_BUF..*\Z
new file mode 100644
@@ -0,0 +1,32 @@
+/*
+ The section .bss.MY_BUF won't fit in RAM1 or RAM2
+*/
+
+MEMORY
+{
+ ROM (rx) : ORIGIN = 0x8000000, LENGTH = 10K
+ RAM1 (xrw) : ORIGIN = 0x10000000, LENGTH = 64K
+ RAM2 (xrw) : ORIGIN = 0x20000000, LENGTH = 96K
+}
+
+SECTIONS
+{
+ .text :
+ {
+ KEEP(*(.text.foo)) ;
+ } >ROM
+
+ .bss :
+ {
+ _sbss = .;
+ *(.bss) *(.bss*) ;
+ _ebss = .;
+ } >RAM1
+
+ .bss_ram2 :
+ {
+ _sbss_ram2 = .;
+ *(.bss) *(.bss*) ;
+ _ebss_ram2 = .;
+ } >RAM2
+}
new file mode 100644
@@ -0,0 +1,16 @@
+ .global MY_BUF
+ .section .bss.MY_BUF,"aw",%nobits
+ .type MY_BUF, %object
+ .size MY_BUF, 102400
+MY_BUF:
+ .space 102400
+
+ .section .text.foo,"ax",%progbits
+ .global foo
+ .type foo, %function
+foo:
+ ldr r0, .L3
+ bx lr
+.L3:
+ .word MY_BUF
+ .size foo, .-foo