[PR22764,LD,AARCH64] Allow R_AARCH64_ABS16 and R_AARCH64_ABS32 against absolution symbol or undefine symbol in shared object under LP64 abi.

Message ID 80514701-2aad-d704-12f2-865073b60a99@foss.arm.com
State New
Headers show
Series
  • [PR22764,LD,AARCH64] Allow R_AARCH64_ABS16 and R_AARCH64_ABS32 against absolution symbol or undefine symbol in shared object under LP64 abi.
Related show

Commit Message

Renlin Li Feb. 5, 2018, 12:39 p.m.
Hi all,

The assumption that R_AARCH64_ABS16 and R_AARCH64_ABS32 relocation in LP64 abi
will be used to generate an address does not hold for absolute symbol.
In this case, it is a value fixed at static linking time.

The condition to check the relocations is relax to allow absolute symbol and
undefined symbol case.

aarch64 binutils cross/native regression test Okay. Linux kernel link Okay.

Okay to commit and backport to 2.30 branch?

Regards,
Renlin

bfd/ChangeLog:

2018-02-03  Renlin Li  <renlin.li@arm.com>

         PR ld/22764
         * elfnn-aarch64.c (elfNN_aarch64_check_relocs): Relax the
         R_AARCH64_ABS32 and R_AARCH64_ABS16 for absolute symbol. Apply the
         check for writable section as well.

ld/ChangeLog:

2018-02-03  Renlin Li  <renlin.li@arm.com>

         PR ld/22764
         * testsuite/ld-aarch64/emit-relocs-258.s: Define symbol as an address.
         * testsuite/ld-aarch64/emit-relocs-259.s: Likewise.
         * ld/testsuite/ld-aarch64/aarch64-elf.exp: Run new test.
         * testsuite/ld-aarch64/pr22764.s: New.
         * testsuite/ld-aarch64/pr22764.d: New.

Comments

Nick Clifton Feb. 5, 2018, 12:53 p.m. | #1
Hi Renlin,

> Okay to commit and backport to 2.30 branch?


Approved for branch and mainline.

One thing though - and I am just being paranoid now -
can you get an indirect symbol that points to an absolute
symbol ?

If so, then you would need to enhance your check to follow
any indirect symbol to see what type it (eventually) becomes...

Cheers
  Nick
Renlin Li Feb. 5, 2018, 2:10 p.m. | #2
Hi Nick,

Thanks for the review!

Regarding for the indirect symbol case you pointed out,
I saw the following code in elfNN_aarch64_check_relocs ()

>       if (r_symndx < symtab_hdr->sh_info)

> 	{

> 	  /* A local symbol.  */

> 	  ...

> 	}

>       else

> 	{

> 	  h = sym_hashes[r_symndx - symtab_hdr->sh_info];

> 	  while (h->root.type == bfd_link_hash_indirect

> 		 || h->root.type == bfd_link_hash_warning)

> 	    h = (struct elf_link_hash_entry *) h->root.u.i.link;

> 	}


This is done before the check happens.
Does this mean, h is already been processed to points to final symbol?

Regards,
Renlin


On 05/02/18 12:53, Nick Clifton wrote:
> Hi Renlin,

> 

>> Okay to commit and backport to 2.30 branch?

> 

> Approved for branch and mainline.

> 

> One thing though - and I am just being paranoid now -

> can you get an indirect symbol that points to an absolute

> symbol ?

> 

> If so, then you would need to enhance your check to follow

> any indirect symbol to see what type it (eventually) becomes...

> 

> Cheers

>    Nick

>
Nick Clifton Feb. 5, 2018, 5:23 p.m. | #3
Hi Renlin,

> Regarding for the indirect symbol case you pointed out,

> I saw the following code in elfNN_aarch64_check_relocs ()

> 

>>       if (r_symndx < symtab_hdr->sh_info)

>>     {

>>       /* A local symbol.  */

>>       ...

>>     }

>>       else

>>     {

>>       h = sym_hashes[r_symndx - symtab_hdr->sh_info];

>>       while (h->root.type == bfd_link_hash_indirect

>>          || h->root.type == bfd_link_hash_warning)

>>         h = (struct elf_link_hash_entry *) h->root.u.i.link;

>>     }

> 

> This is done before the check happens.

> Does this mean, h is already been processed to points to final symbol?


Yes. :-)  [I should have looked further back in the code when I was reviewing the patch].

Cheers
  Nick
Renlin Li Feb. 6, 2018, 2:21 p.m. | #4
Hi Nick,

On 05/02/18 17:23, Nick Clifton wrote:
> Hi Renlin,

> 

>> Regarding for the indirect symbol case you pointed out,

>> I saw the following code in elfNN_aarch64_check_relocs ()

>>

>>>        if (r_symndx < symtab_hdr->sh_info)

>>>      {

>>>        /* A local symbol.  */

>>>        ...

>>>      }

>>>        else

>>>      {

>>>        h = sym_hashes[r_symndx - symtab_hdr->sh_info];

>>>        while (h->root.type == bfd_link_hash_indirect

>>>           || h->root.type == bfd_link_hash_warning)

>>>          h = (struct elf_link_hash_entry *) h->root.u.i.link;

>>>      }

>>

>> This is done before the check happens.

>> Does this mean, h is already been processed to points to final symbol?

> 

> Yes. :-)  [I should have looked further back in the code when I was reviewing the patch].


Not at all! It is a good point. I didn't consider the case you mentioned when I wrote the patch.
It happens that it is properly handled.

Regards,
Renlin


> 

> Cheers

>    Nick

> 

>   

>

Patch

diff --git a/bfd/elfnn-aarch64.c b/bfd/elfnn-aarch64.c
index af448f98aef6fb39081d932a22951b58d8fc6515..2737773d1ccfa33c661f10b99a4937e43b46dfda 100644
--- a/bfd/elfnn-aarch64.c
+++ b/bfd/elfnn-aarch64.c
@@ -7189,10 +7189,19 @@  elfNN_aarch64_check_relocs (bfd *abfd, struct bfd_link_info *info,
 #if ARCH_SIZE == 64
 	case BFD_RELOC_AARCH64_32:
 #endif
-	  if (bfd_link_pic (info)
-	      && (sec->flags & SEC_ALLOC) != 0
-	      && (sec->flags & SEC_READONLY) != 0)
+	  if (bfd_link_pic (info) && (sec->flags & SEC_ALLOC) != 0)
 	    {
+	      if (h != NULL
+		  /* This is an absolute symbol.  It represents a value instead
+		     of an address.  */
+		  && ((h->root.type == bfd_link_hash_defined
+		       && bfd_is_abs_section (h->root.u.def.section))
+		      /* This is an undefined symbol.  */
+		      || h->root.type == bfd_link_hash_undefined))
+		break;
+
+	      /* For local symbols, defined global symbols in a non-ABS section,
+		 it is assumed that the value is an address.  */
 	      int howto_index = bfd_r_type - BFD_RELOC_AARCH64_RELOC_START;
 	      _bfd_error_handler
 		/* xgettext:c-format */
diff --git a/ld/testsuite/ld-aarch64/aarch64-elf.exp b/ld/testsuite/ld-aarch64/aarch64-elf.exp
index 2602a43dd90558e80c0f384abe393b7f41fd850f..c67ffb19ebfa074bdd4d203a5ef11bd0a61cb25e 100644
--- a/ld/testsuite/ld-aarch64/aarch64-elf.exp
+++ b/ld/testsuite/ld-aarch64/aarch64-elf.exp
@@ -285,6 +285,7 @@  run_dump_test "pr17415"
 run_dump_test_lp64 "tprel_g2_overflow"
 run_dump_test "tprel_add_lo12_overflow"
 run_dump_test "protected-data"
+run_dump_test_lp64 "pr22764"
 
 # ifunc tests
 run_dump_test "ifunc-1"
diff --git a/ld/testsuite/ld-aarch64/emit-relocs-258.s b/ld/testsuite/ld-aarch64/emit-relocs-258.s
index f724776243199f4a9ff94b634b0998b963f92cf6..87bb657c5d99e739f7845869c4d221e4a485497a 100644
--- a/ld/testsuite/ld-aarch64/emit-relocs-258.s
+++ b/ld/testsuite/ld-aarch64/emit-relocs-258.s
@@ -1,5 +1,6 @@ 
+.global dummy
 .text
-
+dummy:
   ldr x0, .L1
 
 .L1:
diff --git a/ld/testsuite/ld-aarch64/emit-relocs-259.s b/ld/testsuite/ld-aarch64/emit-relocs-259.s
index 7e1ba3ceb400b582350aa43d3c7d925ef22efd53..0977c9d869f3bfd224c439310d1f0b34073dbcd0 100644
--- a/ld/testsuite/ld-aarch64/emit-relocs-259.s
+++ b/ld/testsuite/ld-aarch64/emit-relocs-259.s
@@ -1,5 +1,6 @@ 
+.global dummy
 .text
-
+dummy:
   ldr x0, .L1
 
 .L1:
diff --git a/ld/testsuite/ld-aarch64/pr22764.d b/ld/testsuite/ld-aarch64/pr22764.d
new file mode 100644
index 0000000000000000000000000000000000000000..46a063dea8565726b6d99c510462b2c1a6d608a3
--- /dev/null
+++ b/ld/testsuite/ld-aarch64/pr22764.d
@@ -0,0 +1,17 @@ 
+#source: pr22764.s
+#ld: -shared -T relocs.ld -defsym sym_abs1=0x1 -defsym sym_abs2=0x2 -defsym sym_abs3=0x3 -e0 --emit-relocs
+#objdump: -dr
+#...
+
+Disassembly of section \.text:
+
+0000000000010000 \<\.text\>:
+   10000:	d503201f 	nop
+	...
+			10004: R_AARCH64_ABS64	sym_abs1
+   1000c:	00000002 	\.word	0x00000002
+			1000c: R_AARCH64_ABS32	sym_abs2
+   10010:	0003      	\.short	0x0003
+			10010: R_AARCH64_ABS16	sym_abs3
+   10012:	0000      	\.short	0x0000
+   10014:	d503201f 	nop
diff --git a/ld/testsuite/ld-aarch64/pr22764.s b/ld/testsuite/ld-aarch64/pr22764.s
new file mode 100644
index 0000000000000000000000000000000000000000..25e36b4a1290b8e84c8c18c55c69bb9f8e8a686b
--- /dev/null
+++ b/ld/testsuite/ld-aarch64/pr22764.s
@@ -0,0 +1,6 @@ 
+  .text
+  nop
+  .xword sym_abs1
+  .word sym_abs2
+  .short sym_abs3
+  nop