* [PATCH] elf: Assert range of ns argument in _dl_debug_initialize
@ 2021-06-23 12:42 Florian Weimer
2021-06-27 22:15 ` Carlos O'Donell
2021-06-28 7:26 ` Andreas Schwab
0 siblings, 2 replies; 6+ messages in thread
From: Florian Weimer @ 2021-06-23 12:42 UTC (permalink / raw)
To: libc-alpha
This does not fix any bugs as such, but makes it more obvious
if _dl_debug_initialize is called with invalid arguments
(which would otherwise cause the function to clobber unrelated
data).
Tested on i686-linux-gnu and x86_64-linux-gnu.
---
elf/dl-debug.c | 8 +++++++-
1 file changed, 7 insertions(+), 1 deletion(-)
diff --git a/elf/dl-debug.c b/elf/dl-debug.c
index 2cd5f09753..85b087455e 100644
--- a/elf/dl-debug.c
+++ b/elf/dl-debug.c
@@ -16,6 +16,8 @@
License along with the GNU C Library; if not, see
<https://www.gnu.org/licenses/>. */
+#include <array_length.h>
+#include <assert.h>
#include <ldsodefs.h>
@@ -49,7 +51,11 @@ _dl_debug_initialize (ElfW(Addr) ldbase, Lmid_t ns)
if (ns == LM_ID_BASE)
r = &_r_debug;
else
- r = &GL(dl_ns)[ns]._ns_debug;
+ {
+ assert (ns >= 0);
+ assert (ns < array_length (GL (dl_ns)));
+ r = &GL(dl_ns)[ns]._ns_debug;
+ }
if (r->r_map == NULL || ldbase != 0)
{
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: [PATCH] elf: Assert range of ns argument in _dl_debug_initialize
2021-06-23 12:42 [PATCH] elf: Assert range of ns argument in _dl_debug_initialize Florian Weimer
@ 2021-06-27 22:15 ` Carlos O'Donell
2021-06-27 22:51 ` Florian Weimer
2021-06-28 7:26 ` Andreas Schwab
1 sibling, 1 reply; 6+ messages in thread
From: Carlos O'Donell @ 2021-06-27 22:15 UTC (permalink / raw)
To: Florian Weimer, libc-alpha
On 6/23/21 8:42 AM, Florian Weimer via Libc-alpha wrote:
> This does not fix any bugs as such, but makes it more obvious
> if _dl_debug_initialize is called with invalid arguments
> (which would otherwise cause the function to clobber unrelated
> data).
>
> Tested on i686-linux-gnu and x86_64-linux-gnu.
I know I'm expanding the scope here to include _dl_map_object,
but it's another place where we have a similar check, and so
I'm just thinking that for consistency we should make both
robust in the same way.
I'm not asking you to fix what appears to be a problem in
dl_open_worker that we appear to do *no* validation of nsid
which is user controlled as input to dlopen :>
> ---
> elf/dl-debug.c | 8 +++++++-
> 1 file changed, 7 insertions(+), 1 deletion(-)
>
> diff --git a/elf/dl-debug.c b/elf/dl-debug.c
> index 2cd5f09753..85b087455e 100644
> --- a/elf/dl-debug.c
> +++ b/elf/dl-debug.c
> @@ -16,6 +16,8 @@
> License along with the GNU C Library; if not, see
> <https://www.gnu.org/licenses/>. */
>
> +#include <array_length.h>
> +#include <assert.h>
> #include <ldsodefs.h>
>
>
> @@ -49,7 +51,11 @@ _dl_debug_initialize (ElfW(Addr) ldbase, Lmid_t ns)
> if (ns == LM_ID_BASE)
> r = &_r_debug;
> else
> - r = &GL(dl_ns)[ns]._ns_debug;
> + {
> + assert (ns >= 0);
> + assert (ns < array_length (GL (dl_ns)));
The check in _dl_map_object is:
assert (nsid >= 0);
assert (nsid < GL(dl_nns));
Should we be consistent one way or the other?
> + r = &GL(dl_ns)[ns]._ns_debug;
> + }
>
> if (r->r_map == NULL || ldbase != 0)
> {
>
--
Cheers,
Carlos.
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: [PATCH] elf: Assert range of ns argument in _dl_debug_initialize
2021-06-27 22:15 ` Carlos O'Donell
@ 2021-06-27 22:51 ` Florian Weimer
2021-06-28 2:15 ` Carlos O'Donell
0 siblings, 1 reply; 6+ messages in thread
From: Florian Weimer @ 2021-06-27 22:51 UTC (permalink / raw)
To: Carlos O'Donell; +Cc: libc-alpha
* Carlos O'Donell:
>> diff --git a/elf/dl-debug.c b/elf/dl-debug.c
>> index 2cd5f09753..85b087455e 100644
>> --- a/elf/dl-debug.c
>> +++ b/elf/dl-debug.c
>> @@ -16,6 +16,8 @@
>> License along with the GNU C Library; if not, see
>> <https://www.gnu.org/licenses/>. */
>>
>> +#include <array_length.h>
>> +#include <assert.h>
>> #include <ldsodefs.h>
>>
>>
>> @@ -49,7 +51,11 @@ _dl_debug_initialize (ElfW(Addr) ldbase, Lmid_t ns)
>> if (ns == LM_ID_BASE)
>> r = &_r_debug;
>> else
>> - r = &GL(dl_ns)[ns]._ns_debug;
>> + {
>> + assert (ns >= 0);
>> + assert (ns < array_length (GL (dl_ns)));
>
> The check in _dl_map_object is:
> assert (nsid >= 0);
> assert (nsid < GL(dl_nns));
>
> Should we be consistent one way or the other?
I wasn't sure if _dl_debug_initialize can be called with a
not-yet-allocated (or already-deallocated) namespace ID. _dl_map_object
is somewhat higher-level, so it's not surprising that it expects an
active ID. An out-of-bounds array access is clearly invalid, though.
Thanks,
Florian
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: [PATCH] elf: Assert range of ns argument in _dl_debug_initialize
2021-06-27 22:51 ` Florian Weimer
@ 2021-06-28 2:15 ` Carlos O'Donell
2021-06-28 8:31 ` Florian Weimer
0 siblings, 1 reply; 6+ messages in thread
From: Carlos O'Donell @ 2021-06-28 2:15 UTC (permalink / raw)
To: Florian Weimer; +Cc: libc-alpha
On 6/27/21 6:51 PM, Florian Weimer wrote:
> * Carlos O'Donell:
>
>>> diff --git a/elf/dl-debug.c b/elf/dl-debug.c
>>> index 2cd5f09753..85b087455e 100644
>>> --- a/elf/dl-debug.c
>>> +++ b/elf/dl-debug.c
>>> @@ -16,6 +16,8 @@
>>> License along with the GNU C Library; if not, see
>>> <https://www.gnu.org/licenses/>. */
>>>
>>> +#include <array_length.h>
>>> +#include <assert.h>
>>> #include <ldsodefs.h>
>>>
>>>
>>> @@ -49,7 +51,11 @@ _dl_debug_initialize (ElfW(Addr) ldbase, Lmid_t ns)
>>> if (ns == LM_ID_BASE)
>>> r = &_r_debug;
>>> else
>>> - r = &GL(dl_ns)[ns]._ns_debug;
>>> + {
>>> + assert (ns >= 0);
>>> + assert (ns < array_length (GL (dl_ns)));
>>
>> The check in _dl_map_object is:
>> assert (nsid >= 0);
>> assert (nsid < GL(dl_nns));
>>
>> Should we be consistent one way or the other?
>
> I wasn't sure if _dl_debug_initialize can be called with a
> not-yet-allocated (or already-deallocated) namespace ID. _dl_map_object
> is somewhat higher-level, so it's not surprising that it expects an
> active ID. An out-of-bounds array access is clearly invalid, though.
Assert on the tighter bound and we'll see? :-)
--
Cheers,
Carlos.
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: [PATCH] elf: Assert range of ns argument in _dl_debug_initialize
2021-06-23 12:42 [PATCH] elf: Assert range of ns argument in _dl_debug_initialize Florian Weimer
2021-06-27 22:15 ` Carlos O'Donell
@ 2021-06-28 7:26 ` Andreas Schwab
1 sibling, 0 replies; 6+ messages in thread
From: Andreas Schwab @ 2021-06-28 7:26 UTC (permalink / raw)
To: Florian Weimer via Libc-alpha; +Cc: Florian Weimer
On Jun 23 2021, Florian Weimer via Libc-alpha wrote:
> @@ -49,7 +51,11 @@ _dl_debug_initialize (ElfW(Addr) ldbase, Lmid_t ns)
> if (ns == LM_ID_BASE)
> r = &_r_debug;
> else
> - r = &GL(dl_ns)[ns]._ns_debug;
> + {
> + assert (ns >= 0);
> + assert (ns < array_length (GL (dl_ns)));
The asserts can be combined.
Andreas.
--
Andreas Schwab, schwab@linux-m68k.org
GPG Key fingerprint = 7578 EB47 D4E5 4D69 2510 2552 DF73 E780 A9DA AEC1
"And now for something completely different."
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: [PATCH] elf: Assert range of ns argument in _dl_debug_initialize
2021-06-28 2:15 ` Carlos O'Donell
@ 2021-06-28 8:31 ` Florian Weimer
0 siblings, 0 replies; 6+ messages in thread
From: Florian Weimer @ 2021-06-28 8:31 UTC (permalink / raw)
To: Carlos O'Donell; +Cc: libc-alpha
* Carlos O'Donell:
> On 6/27/21 6:51 PM, Florian Weimer wrote:
>> * Carlos O'Donell:
>>
>>>> diff --git a/elf/dl-debug.c b/elf/dl-debug.c
>>>> index 2cd5f09753..85b087455e 100644
>>>> --- a/elf/dl-debug.c
>>>> +++ b/elf/dl-debug.c
>>>> @@ -16,6 +16,8 @@
>>>> License along with the GNU C Library; if not, see
>>>> <https://www.gnu.org/licenses/>. */
>>>>
>>>> +#include <array_length.h>
>>>> +#include <assert.h>
>>>> #include <ldsodefs.h>
>>>>
>>>>
>>>> @@ -49,7 +51,11 @@ _dl_debug_initialize (ElfW(Addr) ldbase, Lmid_t ns)
>>>> if (ns == LM_ID_BASE)
>>>> r = &_r_debug;
>>>> else
>>>> - r = &GL(dl_ns)[ns]._ns_debug;
>>>> + {
>>>> + assert (ns >= 0);
>>>> + assert (ns < array_length (GL (dl_ns)));
>>>
>>> The check in _dl_map_object is:
>>> assert (nsid >= 0);
>>> assert (nsid < GL(dl_nns));
>>>
>>> Should we be consistent one way or the other?
>>
>> I wasn't sure if _dl_debug_initialize can be called with a
>> not-yet-allocated (or already-deallocated) namespace ID. _dl_map_object
>> is somewhat higher-level, so it's not surprising that it expects an
>> active ID. An out-of-bounds array access is clearly invalid, though.
>
> Assert on the tighter bound and we'll see? :-)
Do we do things this way?
I'm mainly interested in catching the LM_ID_NEWLM case, to be honest.
Thanks,
Florian
^ permalink raw reply [flat|nested] 6+ messages in thread
end of thread, other threads:[~2021-06-28 8:31 UTC | newest]
Thread overview: 6+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2021-06-23 12:42 [PATCH] elf: Assert range of ns argument in _dl_debug_initialize Florian Weimer
2021-06-27 22:15 ` Carlos O'Donell
2021-06-27 22:51 ` Florian Weimer
2021-06-28 2:15 ` Carlos O'Donell
2021-06-28 8:31 ` Florian Weimer
2021-06-28 7:26 ` Andreas Schwab
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).