Received-SPF: pass (google.com: domain of linux-kernel+bounces-186794-linux.lists.archive=gmail.com@vger.kernel.org designates 2604:1380:45d1:ec00::1 as permitted sender) client-ip=2604:1380:45d1:ec00::1;
From: Yury Norov <yury.norov@gmail.com>
To: linux-kernel@vger.kernel.org
Cc: Yury Norov <yury.norov@gmail.com>,
	Petr Mladek <pmladek@suse.com>,
	Andrew Morton <akpm@linux-foundation.org>,
	Masahiro Yamada <masahiroy@kernel.org>,
	Randy Dunlap <rdunlap@infradead.org>,
	Yoann Congal <yoann.congal@smile.fr>,
	"Gustavo A. R. Silva" <gustavoars@kernel.org>,
	Nhat Pham <nphamcs@gmail.com>,
	Vincent Guittot <vincent.guittot@linaro.org>,
	Alexander Lobakin <aleksander.lobakin@intel.com>,
	"David S . Miller" <davem@davemloft.net>
Subject: [PATCH] gcc: disable '-Warray-bounds' for gcc-9
Date: Wed, 22 May 2024 15:58:30 -0700
Message-Id: <20240522225830.1201778-1-yury.norov@gmail.com>
Precedence: bulk
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

'-Warray-bounds' is already disabled for gcc-10+. Now that we've merged
bitmap_{read,write), I see the following error when building the kernel
with gcc-9.4 (Ubuntu 20.04.4 LTS) for x86_64 allmodconfig:

drivers/pinctrl/pinctrl-cy8c95x0.c: In function ‘cy8c95x0_read_regs_mask.isra.0’:
include/linux/bitmap.h:756:18: error: array subscript [1, 288230376151711744] is outside array bounds of ‘long unsigned int[1]’ [-Werror=array-bounds]
  756 |  value_high = map[index + 1] & BITMAP_LAST_WORD_MASK(start + nbits);
      |               ~~~^~~~~~~~~~~

The immediate reason is that the commit b44759705f7d ("bitmap: make
bitmap_{get,set}_value8() use bitmap_{read,write}()") switched the
bitmap_get_value8() to an alias of bitmap_read(); the same for 'set'.

Now; the code that triggers Warray-bounds, calls the function like this:

  #define MAX_BANK 8
  #define BANK_SZ 8
  #define MAX_LINE        (MAX_BANK * BANK_SZ)
  DECLARE_BITMAP(tval, MAX_LINE); // 64-bit map: unsigned long tval[1]
  
  read_val |= bitmap_get_value8(tval, i * BANK_SZ) & ~bits;

bitmap_read() is implemented such that it may conditionally dereference
a pointer beyond the boundary like this:

	unsigned long offset = start % BITS_PER_LONG;
        unsigned long space = BITS_PER_LONG - offset;

        if (space >= nbits)
                return (map[index] >> offset) & BITMAP_LAST_WORD_MASK(nbits);

        value_low = map[index] & BITMAP_FIRST_WORD_MASK(start);
        value_high = map[index + 1] & BITMAP_LAST_WORD_MASK(start + nbits);
        return (value_low >> offset) | (value_high << space);

In case of bitmap_get_value8(), it's impossible to violate the boundary
because 'space >= nbits' is never the true for byte-aligned 8-bit access.
So, this is clearly a false-positive. 

The same type of false-positives break my allmodconfig build in many
places. gcc-8, is clear, however.

Signed-off-by: Yury Norov <yury.norov@gmail.com>

---
It has already been spotted in this thread:

https://lore.kernel.org/linux-kernel//457b9cbb-9a5f-47ef-9eac-3e4f135d6a96@intel.com/T/#mc833ceac1b05d7156a864dcfe21435d01d77a0d7

And I suggested to use the OPTIMIZER_HIDE_VAR() there. But now that I
see the warning disabled for gcc-10+, it looks simpler to just extend
it to gcc-9+.

---
 init/Kconfig | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/init/Kconfig b/init/Kconfig
index 72404c1f2157..febdea2afc3b 100644
--- a/init/Kconfig
+++ b/init/Kconfig
@@ -883,7 +883,7 @@ config GCC10_NO_ARRAY_BOUNDS
 
 config CC_NO_ARRAY_BOUNDS
 	bool
-	default y if CC_IS_GCC && GCC_VERSION >= 100000 && GCC10_NO_ARRAY_BOUNDS
+	default y if CC_IS_GCC && GCC_VERSION >= 90000 && GCC10_NO_ARRAY_BOUNDS
 
 # Currently, disable -Wstringop-overflow for GCC globally.
 config GCC_NO_STRINGOP_OVERFLOW
-- 
2.40.1