public inbox for gcc-patches@gcc.gnu.org
 help / color / mirror / Atom feed
From: Martin Sebor <msebor@gmail.com>
To: Thomas Schwinge <thomas@codesourcery.com>
Cc: gcc-patches@gcc.gnu.org,
	Christophe Lyon <christophe.lyon@linaro.org>,
	Hafiz Abid Qadeer <abidh@codesourcery.com>,
	Andrew Stubbs <ams@codesourcery.com>
Subject: Re: [PATCH libatomic/arm] avoid warning on constant addresses (PR 101379)
Date: Fri, 16 Jul 2021 15:11:24 -0600	[thread overview]
Message-ID: <1ab598da-6cac-c1bd-b54f-4f5c8b0683f6@gmail.com> (raw)
In-Reply-To: <87pmvii29w.fsf@euler.schwinge.homeip.net>

On 7/16/21 11:42 AM, Thomas Schwinge wrote:
> Hi Martin!
> 
> On 2021-07-09T17:11:25-0600, Martin Sebor via Gcc-patches <gcc-patches@gcc.gnu.org> wrote:
>> The attached tweak avoids the new -Warray-bounds instances when
>> building libatomic for arm. Christophe confirms it resolves
>> the problem (thank you!)
> 
> As Abid has just reported in
> <https://gcc.gnu.org/bugzilla/show_bug.cgi?id=101374#c16>, similar
> problem with GCN target libgomp build:
> 
>      In function ‘gcn_thrs’,
>          inlined from ‘gomp_thread’ at [...]/source-gcc/libgomp/libgomp.h:803:10,
>          inlined from ‘GOMP_barrier’ at [...]/source-gcc/libgomp/barrier.c:34:29:
>      [...]/source-gcc/libgomp/libgomp.h:792:10: error: array subscript 0 is outside array bounds of ‘__lds struct gomp_thread * __lds[0]’ [-Werror=array-bounds]
>        792 |   return *thrs;
>            |          ^~~~~
> 
>      gcc/config/gcn/gcn.h:  c_register_addr_space ("__lds", ADDR_SPACE_LDS);                   \
> 
>      libgomp/libgomp.h-static inline struct gomp_thread *gcn_thrs (void)
>      libgomp/libgomp.h-{
>      libgomp/libgomp.h-  /* The value is at the bottom of LDS.  */
>      libgomp/libgomp.h:  struct gomp_thread * __lds *thrs = (struct gomp_thread * __lds *)4;
>      libgomp/libgomp.h-  return *thrs;
>      libgomp/libgomp.h-}
> 
> ..., plus a few more.  Work-around:
> 
>         struct gomp_thread * __lds *thrs = (struct gomp_thread * __lds *)4;
>      +# pragma GCC diagnostic push
>      +# pragma GCC diagnostic ignored "-Warray-bounds"
>         return *thrs;
>      +# pragma GCC diagnostic pop
> 
> ..., but it's a bit tedious to add that in all that the other places,
> too.  (So I'll consider some GCN-specific '-Wno-array-bounds' if we don't
> get to resolve this otherwise, soon.)
> 
>> As we have discussed, the main goal of this class of warnings
>> is to detect accesses at addresses derived from null pointers
>> (e.g., to struct members or array elements at a nonzero offset).
> 
> (ACK, and thanks for that work!)
> 
>> Diagnosing accesses at hardcoded addresses is incidental because
>> at the stage they are detected the two are not distinguishable
>> from each another.
>>
>> I'm planning (hoping) to implement detection of invalid pointer
>> arithmetic involving null for GCC 12, so this patch is a stopgap
>> solution to unblock the arm libatomic build without compromising
>> the warning.  Once the new detection is in place these workarounds
>> can be removed or replaced with something more appropriate (e.g.,
>> declaring the objects at the hardwired addresses with an attribute
>> like AVR's address or io; that would enable bounds checking at
>> those addresses as well).
> 
> Of course, we may simply re-work the libgomp/GCN code -- but don't we
> first need to answer the question whether the current code is actually
> "bad"?  Aren't we going to get a lot of similar reports from
> kernel/embedded/other low-level software developers, once this is out in
> the wild?  I mean:
> 
>> PR bootstrap/101379 - libatomic arm build failure after r12-2132 due to -Warray-bounds on a constant address
>>
>> libatomic/ChangeLog:
>>        * /config/linux/arm/host-config.h (__kernel_helper_version): New
>>        function.  Adjust shadow macro.
>>
>> diff --git a/libatomic/config/linux/arm/host-config.h b/libatomic/config/linux/arm/host-config.h
>> index 1520f237d73..777d08a2b85 100644
>> --- a/libatomic/config/linux/arm/host-config.h
>> +++ b/libatomic/config/linux/arm/host-config.h
>> @@ -39,8 +39,14 @@ typedef void (__kernel_dmb_t) (void);
>>   #define __kernel_dmb (*(__kernel_dmb_t *) 0xffff0fa0)
>>
>>   /* Kernel helper page version number.  */
>> -#define __kernel_helper_version (*(unsigned int *)0xffff0ffc)
> 
> Are such (not un-common) '#define's actually "bad", and anyhow ought to
> be replaced by something like the following?

Like all warnings (and especially flow-based ones that depend on
optimization), this one too involves a trade-off between noise and
real bugs.  There clearly is some low-level code that intentionally
accesses memory at hardcoded addresses.  But because null pointers
are pervasive, there's a lot more code that could end up accessing
data at some offset from zero by accident (e.g., by writing to
an array element or a member of a struct).  This affects all code,
but is an especially big concern for privileged code that can access
all memory.  So in my view, the trade-off is worthwhile.

The logic the warning relies on isn't new: it was introduced in GCC
11.  There have been a handful of reports of this issue (some from
the kernel) but far fewer than in other warnings.  The recent change
expose more code to the logic so the numbers of both false and true
positives are bound to go up, in proportion.  Hopefully, before GCC
12 is released, I will have a more robust solution to the null+offset
problem.

> 
>> +static inline unsigned*
>> +__kernel_helper_version ()
>> +{
>> +  unsigned *volatile addr = (unsigned int *)0xffff0ffc;
>> +  return addr;
>> +}
>>
>> +#define __kernel_helper_version (*__kernel_helper_version())
> 
> (No 'volatile' in the original code, by the way.)

The volatile is what prevents the warning.  But I think a better
solution than the hack above is to introduce a named extern const
variable for the address.  It avoids the issue without the penalty
of multiple volatile accesses and if/when an attribute like AVR
address is introduced it can be more easily adapted to it.  Real
object declarations with an attribute is also a more appropriate
mechanism than using hardcoded address in pointers.

Martin

  reply	other threads:[~2021-07-16 21:11 UTC|newest]

Thread overview: 15+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2021-07-09 23:11 Martin Sebor
2021-07-15  8:33 ` Christophe Lyon
2021-07-16 17:42 ` Thomas Schwinge
2021-07-16 21:11   ` Martin Sebor [this message]
2021-07-19  8:49     ` Thomas Schwinge
2021-07-17 22:28   ` Andrew Stubbs
2021-07-19  8:46     ` [gcn] Work-around libgomp 'error: array subscript 0 is outside array bounds of ‘__lds struct gomp_thread * __lds[0]’ [-Werror=array-bounds]' (was: [PATCH libatomic/arm] avoid warning on constant addresses (PR 101379)) Thomas Schwinge
2021-07-19  8:56       ` Jakub Jelinek
2021-07-19 11:10       ` Andrew Stubbs
2021-07-20  7:23       ` [gcn] Work-around libgomp 'error: array subscript 0 is outside array bounds of ‘__lds struct gomp_thread * __lds[0]’ [-Werror=array-bounds]' Thomas Schwinge
2021-07-20  8:40         ` '#pragma GCC diagnostic' (mis-)use in 'statement' of 'if' (was: [gcn] Work-around libgomp 'error: array subscript 0 is outside array bounds of ‘__lds struct gomp_thread * __lds[0]’ [-Werror=array-bounds]') Thomas Schwinge
2021-07-20 19:47           ` '#pragma GCC diagnostic' (mis-)use in 'statement' of 'if' Martin Sebor
2021-07-20 20:16             ` Jakub Jelinek
2021-07-21 16:41 ` [PATCH libatomic/arm] avoid warning on constant addresses (PR 101379) Kyrylo Tkachov
2021-07-21 16:54   ` Martin Sebor

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=1ab598da-6cac-c1bd-b54f-4f5c8b0683f6@gmail.com \
    --to=msebor@gmail.com \
    --cc=abidh@codesourcery.com \
    --cc=ams@codesourcery.com \
    --cc=christophe.lyon@linaro.org \
    --cc=gcc-patches@gcc.gnu.org \
    --cc=thomas@codesourcery.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
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).