[Patch] ld: check for address space overflow

[Patch] ld: check for address space overflow

[Tristan Gingold]

Hello,

currently the linker allows to put a section which starts near the end of the address space and which ends past the end (so wrap around).  I think this should be flagged as an error.

No regression on powerpc-elf

Ok for master ?

Tristan.

commit 09718888d36a07cc9c71aa301a39a4299dbf8c34
Author: Tristan Gingold <gingold@adacore.com>
Date:   Fri Mar 10 15:16:19 2017 +0100

    ld: add an error in case of address space overflow.
    
    ld/
    	* ldlang.c (lang_check_section_addresses): Check for address space
    	overflow.
    	* testsuite/ld-checks/checks.exp (overflow_check): New procedure
    	* testsuite/ld-checks/over.s: New test source.
    	* testsuite/ld-checks/over.d: New test.
    	* testsuite/ld-checks/over2.s: New test source.
    	* testsuite/ld-checks/over2.d: New test.

diff --git a/ld/ChangeLog b/ld/ChangeLog
index 3883bcb..5387210 100644
--- a/ld/ChangeLog
+++ b/ld/ChangeLog
@@ -1,3 +1,13 @@
+2017-03-10  Tristan Gingold  <gingold@adacore.com>
+
+	* ldlang.c (lang_check_section_addresses): Check for address space
+	overflow.
+	* testsuite/ld-checks/checks.exp (overflow_check): New procedure
+	* testsuite/ld-checks/over.s: New test source.
+	* testsuite/ld-checks/over.d: New test.
+	* testsuite/ld-checks/over2.s: New test source.
+	* testsuite/ld-checks/over2.d: New test.
+
 2017-03-07  Alan Modra  <amodra@gmail.com>
 
 	* ldlang.c (open_input_bfds): Check that lang_assignment_statement
diff --git a/ld/ldlang.c b/ld/ldlang.c
index ff6ef39..36b6884 100644
--- a/ld/ldlang.c
+++ b/ld/ldlang.c
@@ -4768,6 +4768,7 @@ lang_check_section_addresses (void)
   asection *s, *p;
   struct check_sec *sections;
   size_t i, count;
+  bfd_vma addr_mask;
   bfd_vma s_start;
   bfd_vma s_end;
   bfd_vma p_start = 0;
@@ -4775,6 +4776,25 @@ lang_check_section_addresses (void)
   lang_memory_region_type *m;
   bfd_boolean overlays;
 
+  /* Detect address space overflow.  */
+  addr_mask = ((bfd_vma) 1 <<
+	       (bfd_arch_bits_per_address (link_info.output_bfd) - 1)) - 1;
+  addr_mask = (addr_mask << 1) + 1;
+  for (s = link_info.output_bfd->sections; s != NULL; s = s->next)
+    {
+      s_end = (s->vma + s->size) & addr_mask;
+      if (s_end != 0 && s_end < s->vma)
+	einfo (_("%X%P: section %s VMA wraps around address space\n"),
+	       s->name);
+      else
+	{
+	  s_end = (s->lma + s->size) & addr_mask;
+	  if (s_end != 0 && s_end < s->lma)
+	    einfo (_("%X%P: section %s LMA wraps around address space\n"),
+		   s->name);
+	}
+    }
+
   if (bfd_count_sections (link_info.output_bfd) <= 1)
     return;
 
diff --git a/ld/testsuite/ld-checks/checks.exp b/ld/testsuite/ld-checks/checks.exp
index b1c8454..782f50a 100644
--- a/ld/testsuite/ld-checks/checks.exp
+++ b/ld/testsuite/ld-checks/checks.exp
@@ -78,4 +78,17 @@ proc section_check {} {
     }
 }
 
+proc overflow_check {} {
+    # Test only on some 32-bit targets that are often tested
+    if { ![istarget i?86-*-*]
+	 && ![istarget powerpc-*-*]
+	 && ![istarget arm*-*-*] } {
+	return
+    }
+
+    run_dump_test "over"
+    run_dump_test "over2"
+}
+
 section_check
+overflow_check
diff --git a/ld/testsuite/ld-checks/over.d b/ld/testsuite/ld-checks/over.d
new file mode 100644
index 0000000..e34bea9
--- /dev/null
+++ b/ld/testsuite/ld-checks/over.d
@@ -0,0 +1,4 @@
+#name: section size overflow
+#source: over.s
+#ld: -Ttext=0xfffffffc
+#error: .* section .text VMA wraps around address space
diff --git a/ld/testsuite/ld-checks/over.s b/ld/testsuite/ld-checks/over.s
new file mode 100644
index 0000000..7f8d4b5
--- /dev/null
+++ b/ld/testsuite/ld-checks/over.s
@@ -0,0 +1,7 @@
+	.text
+	.globl _start
+_start:
+	.long 0
+	.long 0
+	.long 0
+	.long 0
diff --git a/ld/testsuite/ld-checks/over2.d b/ld/testsuite/ld-checks/over2.d
new file mode 100644
index 0000000..511b917
--- /dev/null
+++ b/ld/testsuite/ld-checks/over2.d
@@ -0,0 +1,8 @@
+#name: section size overflow
+#source: over2.s
+#ld: -Ttext=0xfffffffc
+#nm: -n
+
+#...
+fffffffc T _start
+#pass
diff --git a/ld/testsuite/ld-checks/over2.s b/ld/testsuite/ld-checks/over2.s
new file mode 100644
index 0000000..dc6de0e
--- /dev/null
+++ b/ld/testsuite/ld-checks/over2.s
@@ -0,0 +1,4 @@
+	.text
+	.globl _start
+_start:
+	.long 0

Re: [Patch] ld: check for address space overflow

[Alan Modra]

On Fri, Mar 10, 2017 at 03:29:24PM +0100, Tristan Gingold wrote:
>     ld: add an error in case of address space overflow.
>     
>     ld/
>     	* ldlang.c (lang_check_section_addresses): Check for address space
>     	overflow.
>     	* testsuite/ld-checks/checks.exp (overflow_check): New procedure
>     	* testsuite/ld-checks/over.s: New test source.
>     	* testsuite/ld-checks/over.d: New test.
>     	* testsuite/ld-checks/over2.s: New test source.
>     	* testsuite/ld-checks/over2.d: New test.

OK.

-- 
Alan Modra
Australia Development Lab, IBM

Re: [Patch] ld: check for address space overflow

[Tristan Gingold]

> On 11 Mar 2017, at 02:14, Alan Modra <amodra@gmail.com> wrote:
> 
> On Fri, Mar 10, 2017 at 03:29:24PM +0100, Tristan Gingold wrote:
>>    ld: add an error in case of address space overflow.
>> 
>>    ld/
>>    	* ldlang.c (lang_check_section_addresses): Check for address space
>>    	overflow.
>>    	* testsuite/ld-checks/checks.exp (overflow_check): New procedure
>>    	* testsuite/ld-checks/over.s: New test source.
>>    	* testsuite/ld-checks/over.d: New test.
>>    	* testsuite/ld-checks/over2.s: New test source.
>>    	* testsuite/ld-checks/over2.d: New test.
> 
> OK.

Thanks, just committed.

Tristan.

Re: [Patch] ld: check for address space overflow

[H.J. Lu]

On Mon, Mar 13, 2017 at 2:46 AM, Tristan Gingold <gingold@adacore.com> wrote:
>
>> On 11 Mar 2017, at 02:14, Alan Modra <amodra@gmail.com> wrote:
>>
>> On Fri, Mar 10, 2017 at 03:29:24PM +0100, Tristan Gingold wrote:
>>>    ld: add an error in case of address space overflow.
>>>
>>>    ld/
>>>      * ldlang.c (lang_check_section_addresses): Check for address space
>>>      overflow.
>>>      * testsuite/ld-checks/checks.exp (overflow_check): New procedure
>>>      * testsuite/ld-checks/over.s: New test source.
>>>      * testsuite/ld-checks/over.d: New test.
>>>      * testsuite/ld-checks/over2.s: New test source.
>>>      * testsuite/ld-checks/over2.d: New test.
>>
>> OK.
>
> Thanks, just committed.
>
> Tristan.
>

This caused

FAIL: ld-elf/loadaddr1
FAIL: ld-elf/loadaddr2
FAIL: ld-elf/seg

on Linux/i386 with 64-bit VMA.

-- 
H.J.

Re: [Patch] ld: check for address space overflow

[H.J. Lu]

On Fri, Mar 10, 2017 at 6:29 AM, Tristan Gingold <gingold@adacore.com> wrote:
> Hello,
>
> currently the linker allows to put a section which starts near the end of the address space and which ends past the end (so wrap around).  I think this should be flagged as an error.
>
> No regression on powerpc-elf
>
> Ok for master ?
>
> Tristan.
>
> commit 09718888d36a07cc9c71aa301a39a4299dbf8c34
> Author: Tristan Gingold <gingold@adacore.com>
> Date:   Fri Mar 10 15:16:19 2017 +0100
>
>     ld: add an error in case of address space overflow.
>
>     ld/
>         * ldlang.c (lang_check_section_addresses): Check for address space
>         overflow.
>         * testsuite/ld-checks/checks.exp (overflow_check): New procedure
>         * testsuite/ld-checks/over.s: New test source.
>         * testsuite/ld-checks/over.d: New test.
>         * testsuite/ld-checks/over2.s: New test source.
>         * testsuite/ld-checks/over2.d: New test.
>
> diff --git a/ld/ChangeLog b/ld/ChangeLog
> index 3883bcb..5387210 100644
> --- a/ld/ChangeLog
> +++ b/ld/ChangeLog
> @@ -1,3 +1,13 @@
> +2017-03-10  Tristan Gingold  <gingold@adacore.com>
> +
> +       * ldlang.c (lang_check_section_addresses): Check for address space
> +       overflow.
> +       * testsuite/ld-checks/checks.exp (overflow_check): New procedure
> +       * testsuite/ld-checks/over.s: New test source.
> +       * testsuite/ld-checks/over.d: New test.
> +       * testsuite/ld-checks/over2.s: New test source.
> +       * testsuite/ld-checks/over2.d: New test.
> +
>  2017-03-07  Alan Modra  <amodra@gmail.com>
>
>         * ldlang.c (open_input_bfds): Check that lang_assignment_statement
> diff --git a/ld/ldlang.c b/ld/ldlang.c
> index ff6ef39..36b6884 100644
> --- a/ld/ldlang.c
> +++ b/ld/ldlang.c
> @@ -4768,6 +4768,7 @@ lang_check_section_addresses (void)
>    asection *s, *p;
>    struct check_sec *sections;
>    size_t i, count;
> +  bfd_vma addr_mask;
>    bfd_vma s_start;
>    bfd_vma s_end;
>    bfd_vma p_start = 0;
> @@ -4775,6 +4776,25 @@ lang_check_section_addresses (void)
>    lang_memory_region_type *m;
>    bfd_boolean overlays;
>
> +  /* Detect address space overflow.  */
> +  addr_mask = ((bfd_vma) 1 <<
> +              (bfd_arch_bits_per_address (link_info.output_bfd) - 1)) - 1;

This isn't correct.  See is32bit () in bfd.c.  And it doesn't work
for -0x80000000 with 64-bit VMA for 32-bit targets.

> +  addr_mask = (addr_mask << 1) + 1;
> +  for (s = link_info.output_bfd->sections; s != NULL; s = s->next)
> +    {
> +      s_end = (s->vma + s->size) & addr_mask;
> +      if (s_end != 0 && s_end < s->vma)
> +       einfo (_("%X%P: section %s VMA wraps around address space\n"),
> +              s->name);
> +      else
> +       {
> +         s_end = (s->lma + s->size) & addr_mask;
> +         if (s_end != 0 && s_end < s->lma)
> +           einfo (_("%X%P: section %s LMA wraps around address space\n"),
> +                  s->name);
> +       }
> +    }
> +
>    if (bfd_count_sections (link_info.output_bfd) <= 1)
>      return;

Re: [Patch] ld: check for address space overflow

[H.J. Lu]

On Mon, Mar 13, 2017 at 12:06 PM, H.J. Lu <hjl.tools@gmail.com> wrote:
> On Fri, Mar 10, 2017 at 6:29 AM, Tristan Gingold <gingold@adacore.com> wrote:
>> Hello,
>>
>> currently the linker allows to put a section which starts near the end of the address space and which ends past the end (so wrap around).  I think this should be flagged as an error.
>>
>> No regression on powerpc-elf
>>
>> Ok for master ?
>>
>> Tristan.
>>
>> commit 09718888d36a07cc9c71aa301a39a4299dbf8c34
>> Author: Tristan Gingold <gingold@adacore.com>
>> Date:   Fri Mar 10 15:16:19 2017 +0100
>>
>>     ld: add an error in case of address space overflow.
>>
>>     ld/
>>         * ldlang.c (lang_check_section_addresses): Check for address space
>>         overflow.
>>         * testsuite/ld-checks/checks.exp (overflow_check): New procedure
>>         * testsuite/ld-checks/over.s: New test source.
>>         * testsuite/ld-checks/over.d: New test.
>>         * testsuite/ld-checks/over2.s: New test source.
>>         * testsuite/ld-checks/over2.d: New test.
>>
>> diff --git a/ld/ChangeLog b/ld/ChangeLog
>> index 3883bcb..5387210 100644
>> --- a/ld/ChangeLog
>> +++ b/ld/ChangeLog
>> @@ -1,3 +1,13 @@
>> +2017-03-10  Tristan Gingold  <gingold@adacore.com>
>> +
>> +       * ldlang.c (lang_check_section_addresses): Check for address space
>> +       overflow.
>> +       * testsuite/ld-checks/checks.exp (overflow_check): New procedure
>> +       * testsuite/ld-checks/over.s: New test source.
>> +       * testsuite/ld-checks/over.d: New test.
>> +       * testsuite/ld-checks/over2.s: New test source.
>> +       * testsuite/ld-checks/over2.d: New test.
>> +
>>  2017-03-07  Alan Modra  <amodra@gmail.com>
>>
>>         * ldlang.c (open_input_bfds): Check that lang_assignment_statement
>> diff --git a/ld/ldlang.c b/ld/ldlang.c
>> index ff6ef39..36b6884 100644
>> --- a/ld/ldlang.c
>> +++ b/ld/ldlang.c
>> @@ -4768,6 +4768,7 @@ lang_check_section_addresses (void)
>>    asection *s, *p;
>>    struct check_sec *sections;
>>    size_t i, count;
>> +  bfd_vma addr_mask;
>>    bfd_vma s_start;
>>    bfd_vma s_end;
>>    bfd_vma p_start = 0;
>> @@ -4775,6 +4776,25 @@ lang_check_section_addresses (void)
>>    lang_memory_region_type *m;
>>    bfd_boolean overlays;
>>
>> +  /* Detect address space overflow.  */
>> +  addr_mask = ((bfd_vma) 1 <<
>> +              (bfd_arch_bits_per_address (link_info.output_bfd) - 1)) - 1;
>
> This isn't correct.  See is32bit () in bfd.c.  And it doesn't work
> for -0x80000000 with 64-bit VMA for 32-bit targets.
>
>> +  addr_mask = (addr_mask << 1) + 1;
>> +  for (s = link_info.output_bfd->sections; s != NULL; s = s->next)
>> +    {
>> +      s_end = (s->vma + s->size) & addr_mask;
>> +      if (s_end != 0 && s_end < s->vma)
>> +       einfo (_("%X%P: section %s VMA wraps around address space\n"),
>> +              s->name);
>> +      else
>> +       {
>> +         s_end = (s->lma + s->size) & addr_mask;
>> +         if (s_end != 0 && s_end < s->lma)
>> +           einfo (_("%X%P: section %s LMA wraps around address space\n"),
>> +                  s->name);
>> +       }
>> +    }

This works:

diff --git a/ld/ldlang.c b/ld/ldlang.c
index a0638ea..8c1d829 100644
--- a/ld/ldlang.c
+++ b/ld/ldlang.c
@@ -4783,13 +4783,13 @@ lang_check_section_addresses (void)
   for (s = link_info.output_bfd->sections; s != NULL; s = s->next)
     {
       s_end = (s->vma + s->size) & addr_mask;
-      if (s_end != 0 && s_end < s->vma)
+      if (s_end != 0 && s_end < (s->vma & addr_mask))
   einfo (_("%X%P: section %s VMA wraps around address space\n"),
          s->name);
       else
   {
     s_end = (s->lma + s->size) & addr_mask;
-    if (s_end != 0 && s_end < s->lma)
+    if (s_end != 0 && s_end < (s->lma & addr_mask))
       einfo (_("%X%P: section %s LMA wraps around address space\n"),
         s->name);
   }


>> +
>>    if (bfd_count_sections (link_info.output_bfd) <= 1)
>>      return;



-- 
H.J.

Re: [Patch] ld: check for address space overflow

[Tristan Gingold]

[...]

> This works:
> 
> diff --git a/ld/ldlang.c b/ld/ldlang.c
> index a0638ea..8c1d829 100644
> --- a/ld/ldlang.c
> +++ b/ld/ldlang.c
> @@ -4783,13 +4783,13 @@ lang_check_section_addresses (void)
>   for (s = link_info.output_bfd->sections; s != NULL; s = s->next)
>     {
>       s_end = (s->vma + s->size) & addr_mask;
> -      if (s_end != 0 && s_end < s->vma)
> +      if (s_end != 0 && s_end < (s->vma & addr_mask))
>   einfo (_("%X%P: section %s VMA wraps around address space\n"),
>          s->name);
>       else
>   {
>     s_end = (s->lma + s->size) & addr_mask;
> -    if (s_end != 0 && s_end < s->lma)
> +    if (s_end != 0 && s_end < (s->lma & addr_mask))
>       einfo (_("%X%P: section %s LMA wraps around address space\n"),
>         s->name);
>   }

Yes, that makes sense.
Do you plan to commit it ?

Re: [Patch] ld: check for address space overflow

[Alan Modra]

On Tue, Mar 14, 2017 at 09:15:30AM +0100, Tristan Gingold wrote:
> [...]
> 
> > This works:
> > 
> > diff --git a/ld/ldlang.c b/ld/ldlang.c
> > index a0638ea..8c1d829 100644
> > --- a/ld/ldlang.c
> > +++ b/ld/ldlang.c
> > @@ -4783,13 +4783,13 @@ lang_check_section_addresses (void)
> >   for (s = link_info.output_bfd->sections; s != NULL; s = s->next)
> >     {
> >       s_end = (s->vma + s->size) & addr_mask;
> > -      if (s_end != 0 && s_end < s->vma)
> > +      if (s_end != 0 && s_end < (s->vma & addr_mask))
> >   einfo (_("%X%P: section %s VMA wraps around address space\n"),
> >          s->name);
> >       else
> >   {
> >     s_end = (s->lma + s->size) & addr_mask;
> > -    if (s_end != 0 && s_end < s->lma)
> > +    if (s_end != 0 && s_end < (s->lma & addr_mask))
> >       einfo (_("%X%P: section %s LMA wraps around address space\n"),
> >         s->name);
> >   }
> 
> Yes, that makes sense.
> Do you plan to commit it ?

The new testcases need some tweaking too.  Even after applying the
above I'm left with:

arm-aout  +FAIL: section size overflow
arm-coff  +FAIL: section size overflow
arm-epoc-pe  +FAIL: section size overflow
arm-pe  +FAIL: section size overflow
arm-symbianelf  +FAIL: section size overflow
arm-wince-pe  +FAIL: section size overflow
i386-linuxaout  +FAIL: section size overflow
i386-linuxaout  +FAIL: section size overflow
i386-lynxos  +FAIL: section size overflow
i386-netware  +FAIL: section size overflow
i586-aout  +FAIL: section size overflow
i686-pe  +FAIL: section size overflow

-- 
Alan Modra
Australia Development Lab, IBM

Re: [Patch] ld: check for address space overflow

[Tristan Gingold]

> On 14 Mar 2017, at 11:19, Alan Modra <amodra@gmail.com> wrote:
> 
> On Tue, Mar 14, 2017 at 09:15:30AM +0100, Tristan Gingold wrote:
>> [...]
>> 
>>> This works:
>>> 
>>> diff --git a/ld/ldlang.c b/ld/ldlang.c
>>> index a0638ea..8c1d829 100644
>>> --- a/ld/ldlang.c
>>> +++ b/ld/ldlang.c
>>> @@ -4783,13 +4783,13 @@ lang_check_section_addresses (void)
>>>  for (s = link_info.output_bfd->sections; s != NULL; s = s->next)
>>>    {
>>>      s_end = (s->vma + s->size) & addr_mask;
>>> -      if (s_end != 0 && s_end < s->vma)
>>> +      if (s_end != 0 && s_end < (s->vma & addr_mask))
>>>  einfo (_("%X%P: section %s VMA wraps around address space\n"),
>>>         s->name);
>>>      else
>>>  {
>>>    s_end = (s->lma + s->size) & addr_mask;
>>> -    if (s_end != 0 && s_end < s->lma)
>>> +    if (s_end != 0 && s_end < (s->lma & addr_mask))
>>>      einfo (_("%X%P: section %s LMA wraps around address space\n"),
>>>        s->name);
>>>  }
>> 
>> Yes, that makes sense.
>> Do you plan to commit it ?
> 
> The new testcases need some tweaking too.  Even after applying the
> above I'm left with:
> 
> arm-aout  +FAIL: section size overflow
> arm-coff  +FAIL: section size overflow
> arm-epoc-pe  +FAIL: section size overflow
> arm-pe  +FAIL: section size overflow
> arm-symbianelf  +FAIL: section size overflow
> arm-wince-pe  +FAIL: section size overflow
> i386-linuxaout  +FAIL: section size overflow
> i386-linuxaout  +FAIL: section size overflow
> i386-lynxos  +FAIL: section size overflow
> i386-netware  +FAIL: section size overflow
> i586-aout  +FAIL: section size overflow
> i686-pe  +FAIL: section size overflow

What is the host ?  32 bit or 64 bit ?

Re: [Patch] ld: check for address space overflow

[Alan Modra]

On Tue, Mar 14, 2017 at 11:50:27AM +0100, Tristan Gingold wrote:
> 
> > On 14 Mar 2017, at 11:19, Alan Modra <amodra@gmail.com> wrote:
> > The new testcases need some tweaking too.  Even after applying the
> > above I'm left with:
> > 
> > arm-aout  +FAIL: section size overflow
> > arm-coff  +FAIL: section size overflow
> > arm-epoc-pe  +FAIL: section size overflow
> > arm-pe  +FAIL: section size overflow
> > arm-symbianelf  +FAIL: section size overflow
> > arm-wince-pe  +FAIL: section size overflow
> > i386-linuxaout  +FAIL: section size overflow
> > i386-linuxaout  +FAIL: section size overflow
> > i386-lynxos  +FAIL: section size overflow
> > i386-netware  +FAIL: section size overflow
> > i586-aout  +FAIL: section size overflow
> > i686-pe  +FAIL: section size overflow
> 
> What is the host ?  32 bit or 64 bit ?

64 bit.  x86_64-linux.

-- 
Alan Modra
Australia Development Lab, IBM

Re: [Patch] ld: check for address space overflow

[Tristan Gingold]

> On 14 Mar 2017, at 13:17, Alan Modra <amodra@gmail.com> wrote:
> 
> On Tue, Mar 14, 2017 at 11:50:27AM +0100, Tristan Gingold wrote:
>> 
>>> On 14 Mar 2017, at 11:19, Alan Modra <amodra@gmail.com> wrote:
>>> The new testcases need some tweaking too.  Even after applying the
>>> above I'm left with:
>>> 
>>> arm-aout  +FAIL: section size overflow
>>> arm-coff  +FAIL: section size overflow
>>> arm-epoc-pe  +FAIL: section size overflow
>>> arm-pe  +FAIL: section size overflow
>>> arm-symbianelf  +FAIL: section size overflow
>>> arm-wince-pe  +FAIL: section size overflow
>>> i386-linuxaout  +FAIL: section size overflow
>>> i386-linuxaout  +FAIL: section size overflow
>>> i386-lynxos  +FAIL: section size overflow
>>> i386-netware  +FAIL: section size overflow
>>> i586-aout  +FAIL: section size overflow
>>> i686-pe  +FAIL: section size overflow
>> 
>> What is the host ?  32 bit or 64 bit ?
> 
> 64 bit.  x86_64-linux.

Would that be ok ?
(No need to bother with additional sections).

Tristan.

commit 6e31e599777fb6a9e09f06b2fc957201b2c7eb4f
Author: Tristan Gingold <gingold@adacore.com>
Date:   Tue Mar 14 14:54:37 2017 +0100

    ld-checks: tweak overflow checks.
    
    	* testsuite/ld-checks/checks.exp (overflow_check): Disable for
    	non-elf targets.

diff --git a/ld/ChangeLog b/ld/ChangeLog
index c164710..c988c3b 100644
--- a/ld/ChangeLog
+++ b/ld/ChangeLog
@@ -1,3 +1,8 @@
+2017-03-14  Tristan Gingold  <gingold@adacore.com>
+
+	* testsuite/ld-checks/checks.exp (overflow_check): Disable for
+	non-elf targets.
+
 2017-03-13  Tristan Gingold  <gingold@adacore.com>
 
 	* ldlang.c (lang_check_section_addresses): Check for address space
diff --git a/ld/testsuite/ld-checks/checks.exp b/ld/testsuite/ld-checks/checks.exp
index 782f50a..e411d03 100644
--- a/ld/testsuite/ld-checks/checks.exp
+++ b/ld/testsuite/ld-checks/checks.exp
@@ -85,6 +85,9 @@ proc overflow_check {} {
 	 && ![istarget arm*-*-*] } {
 	return
     }
+    if ![is_elf_format] {
+	return
+    }
 
     run_dump_test "over"
     run_dump_test "over2"

Re: [Patch] ld: check for address space overflow

[H.J. Lu]

On Tue, Mar 14, 2017 at 1:15 AM, Tristan Gingold <gingold@adacore.com> wrote:
> [...]
>
>> This works:
>>
>> diff --git a/ld/ldlang.c b/ld/ldlang.c
>> index a0638ea..8c1d829 100644
>> --- a/ld/ldlang.c
>> +++ b/ld/ldlang.c
>> @@ -4783,13 +4783,13 @@ lang_check_section_addresses (void)
>>   for (s = link_info.output_bfd->sections; s != NULL; s = s->next)
>>     {
>>       s_end = (s->vma + s->size) & addr_mask;
>> -      if (s_end != 0 && s_end < s->vma)
>> +      if (s_end != 0 && s_end < (s->vma & addr_mask))
>>   einfo (_("%X%P: section %s VMA wraps around address space\n"),
>>          s->name);
>>       else
>>   {
>>     s_end = (s->lma + s->size) & addr_mask;
>> -    if (s_end != 0 && s_end < s->lma)
>> +    if (s_end != 0 && s_end < (s->lma & addr_mask))
>>       einfo (_("%X%P: section %s LMA wraps around address space\n"),
>>         s->name);
>>   }
>
> Yes, that makes sense.
> Do you plan to commit it ?
>

This is what I checked in.

-- 
H.J.
---
From 9216a6f33592c350ad50696d5571c82e47b71a5e Mon Sep 17 00:00:00 2001
From: "H.J. Lu" <hjl.tools@gmail.com>
Date: Tue, 14 Mar 2017 09:09:54 -0700
Subject: [PATCH] Use addr_mask to check VMA and LMA

Since BFD64 may be used on 32-bit address, we need to apply addr_mask
to check VMA and LMA.

* ldlang.c (lang_check_section_addresses): Use addr_mask to
check VMA and LMA.
---
 ld/ChangeLog | 5 +++++
 ld/ldlang.c  | 4 ++--
 2 files changed, 7 insertions(+), 2 deletions(-)

diff --git a/ld/ChangeLog b/ld/ChangeLog
index eb0c309..df8bb32 100644
--- a/ld/ChangeLog
+++ b/ld/ChangeLog
@@ -1,3 +1,8 @@
+2017-03-14  H.J. Lu  <hongjiu.lu@intel.com>
+
+ * ldlang.c (lang_check_section_addresses): Use addr_mask to
+ check VMA and LMA.
+
 2017-03-13  Nick Clifton  <nickc@redhat.com>

  PR binutils/21202
diff --git a/ld/ldlang.c b/ld/ldlang.c
index a0638ea..8c1d829 100644
--- a/ld/ldlang.c
+++ b/ld/ldlang.c
@@ -4783,13 +4783,13 @@ lang_check_section_addresses (void)
   for (s = link_info.output_bfd->sections; s != NULL; s = s->next)
     {
       s_end = (s->vma + s->size) & addr_mask;
-      if (s_end != 0 && s_end < s->vma)
+      if (s_end != 0 && s_end < (s->vma & addr_mask))
  einfo (_("%X%P: section %s VMA wraps around address space\n"),
        s->name);
       else
  {
   s_end = (s->lma + s->size) & addr_mask;
-  if (s_end != 0 && s_end < s->lma)
+  if (s_end != 0 && s_end < (s->lma & addr_mask))
     einfo (_("%X%P: section %s LMA wraps around address space\n"),
    s->name);
  }
-- 
2.9.3

Re: [Patch] ld: check for address space overflow

[Alan Modra]

On Tue, Mar 14, 2017 at 02:56:18PM +0100, Tristan Gingold wrote:
>     	* testsuite/ld-checks/checks.exp (overflow_check): Disable for
>     	non-elf targets.

OK.  I'm going to commit the following too.  h8300 and ip2k targets
use a 16-bit address cpu by default, so hit the address wrap check
(more than once) on the 64k section test.

	* testsuite/ld-elf/sec64k.exp: Don't run on h8300 and ip2k.

diff --git a/ld/testsuite/ld-elf/sec64k.exp b/ld/testsuite/ld-elf/sec64k.exp
index 249a909..3bf0c52 100644
--- a/ld/testsuite/ld-elf/sec64k.exp
+++ b/ld/testsuite/ld-elf/sec64k.exp
@@ -36,8 +36,10 @@ if { [istarget "d30v-*-*"]
     return
 }
 
-# ft32, m68hc11, m68hc12 and xgate run out of address space.
-if {   [istarget "ft32-*-*"]
+# ft32, h8300, ip2k, m68hc11, m68hc12 and xgate run out of address space.
+if {[istarget "ft32-*-*"]
+    || [istarget "h8300-*-*"]
+    || [istarget "ip2k-*-*"]
     || [istarget "m68hc1*-*"]
     || [istarget "xgate-*"] } {
     return

-- 
Alan Modra
Australia Development Lab, IBM

Re: [Patch] ld: check for address space overflow

[Tristan Gingold]

> On 15 Mar 2017, at 02:11, Alan Modra <amodra@gmail.com> wrote:
> 
> On Tue, Mar 14, 2017 at 02:56:18PM +0100, Tristan Gingold wrote:
>>    	* testsuite/ld-checks/checks.exp (overflow_check): Disable for
>>    	non-elf targets.
> 
> OK.

Thanks, committed.

>  I'm going to commit the following too.  h8300 and ip2k targets
> use a 16-bit address cpu by default, so hit the address wrap check
> (more than once) on the 64k section test.
> 
> 	* testsuite/ld-elf/sec64k.exp: Don't run on h8300 and ip2k.

Thank you for the head-up.
Maybe the overflow check should also check that the length is not larger than the address space (otherwise it could have missed these h8300 and ip2k cases), but I think this is a too artificial case.

Tristan.