Hi Morimoto-san, Karl,
On Wed, Dec 18, 2019 at 6:22 AM Kuninori Morimoto kuninori.morimoto.gx@renesas.com wrote:
From: Kuninori Morimoto kuninori.morimoto.gx@renesas.com
Current SH will get below warning at strncpy()
In file included from ${LINUX}/arch/sh/include/asm/string.h:3, from ${LINUX}/include/linux/string.h:20, from ${LINUX}/include/linux/bitmap.h:9, from ${LINUX}/include/linux/nodemask.h:95, from ${LINUX}/include/linux/mmzone.h:17, from ${LINUX}/include/linux/gfp.h:6, from ${LINUX}/innclude/linux/slab.h:15, from ${LINUX}/linux/drivers/mmc/host/vub300.c:38: ${LINUX}/drivers/mmc/host/vub300.c: In function 'new_system_port_status': ${LINUX}/arch/sh/include/asm/string_32.h:51:42: warning: array subscript\ 80 is above array bounds of 'char[26]' [-Warray-bounds] : "0" (__dest), "1" (__src), "r" (__src+__n) ~~~~~^~~~
In general, strncpy() should behave like below.
char dest[10]; char *src = "12345"; strncpy(dest, src, 10); // dest = {'1', '2', '3', '4', '5', '\0','\0','\0','\0','\0'}
But, current SH strnpy() has 2 issues. 1st is it will access to out-of-memory (= src + 10).
I believe this is not correct: the code does not really access memory beyond the end of the source string. (Recent) gcc just thinks so, because "__src+__n" is used as a parameter to the routine.
2nd is it needs big fixup for it, and maintenance __asm__ code is difficult.
Yeah, the padding is missing.
To solve these issues, this patch simply uses generic strncpy() instead of architecture specific one.
That will definitely fix the issue, as we assume the generic implementation is correct ;-)
Now, I've just tried, naively, to enable CONFIG_STRING_SELFTEST=y in my rts7751r2d build (without your patch), and boot it in qemu:
String selftests succeeded
Woops, turns out lib/test_string.c does not have any testcases for strncpy()...
So adding test code for the corner cases may be a valuable contribution. Thanks!
Gr{oetje,eeting}s,
Geert
-- Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@linux-m68k.org
In personal conversations with technical people, I call myself a hacker. But when I'm talking to journalists I just say "programmer" or something like that. -- Linus Torvalds
Hi Geert
Thank you for your review
But, current SH strnpy() has 2 issues. 1st is it will access to out-of-memory (= src + 10).
I believe this is not correct: the code does not really access memory beyond the end of the source string. (Recent) gcc just thinks so, because "__src+__n" is used as a parameter to the routine.
I see. Will fix in v2
Thank you for your help !! Best regards --- Kuninori Morimoto
linux-kselftest-mirror@lists.linaro.org