C - Problem about mmap()!

Author

Problem about mmap()!

Hao Xu

2004-08-24, 3:55 am

Hi everyone!

I found that if you want to write to the memory got by mmap(), you have
to get the file descriptor for mmap() in O_RDWR mode. If you got the
file descriptor in O_WRONLY mode, then writing to the memory got by
mmap() will lead to segmentation fault. Anyone knows why? Is this a rule
or a bug?

What if I just want to write to the file and nothing else?

Please, look at my tiny program, I wrote this to test mmap(), it is cute.

#include <stdio.h>

typedef struct{
char name[4];
int age;
} people;

int main(int argc, char** argv) {
int fd, i;
people *p_map;
char temp;

fd = open(argv[1], O_RDWR|O_TRUNC); //this works fine

//fd = open(argv[1], O_WRONLY|O_TRUNC); this will cause segmentation fault

if (fd == -1) {
perror("open()");
}
ls

(fd, sizeof(people)*5-1, SEEK_SET);
write(fd, " " , 1);

p_map = (people *)mmap(NULL,sizeof(people)*5,PROT_WRITE,
MAP_SHARED,fd,0);
close(fd);

memset(p_map, '\0', sizeof(people)*5);

temp = 'b';
for(i=0; i<5; i++) {
temp++;
//(p_map+i)->name = "aaa";
//memcpy((p_map+i)->name, &temp, 2);
memset((p_map+i)->name, temp, 3);
(p_map+i)->age = 20+i;
}
printf("initialize over\n");
for(i=0; i<5; i++) {
printf("%s\t%d\n", (p_map+i)->name, (p_map+i)->age);
}
sleep(10);

munmap(p_map, sizeof(people)*5);
printf( "umap ok\n" );

return 0;
}

Fredrik Tolf

2004-08-24, 3:55 am

On Tue, 2004-08-24 at 09:52 +0800, Hao Xu wrote:
> Hi everyone!
>
> I found that if you want to write to the memory got by mmap(), you have
> to get the file descriptor for mmap() in O_RDWR mode. If you got the
> file descriptor in O_WRONLY mode, then writing to the memory got by
> mmap() will lead to segmentation fault. Anyone knows why? Is this a rule
> or a bug?

That is highly implementation-defined. You would probably be much more
likely to get a sensible reply if you post to the newsgroup dealing with
your operating system and/or hardware platform.

Fredrik Tolf

junky_fellow

2004-08-24, 9:02 am

Hao Xu <traverser@vip.sina.com> wrote in message news:<cge70r$2jik$1@mail.cn99.com>...
> Hi everyone!
>
> I found that if you want to write to the memory got by mmap(), you have
> to get the file descriptor for mmap() in O_RDWR mode. If you got the
> file descriptor in O_WRONLY mode, then writing to the memory got by
> mmap() will lead to segmentation fault. Anyone knows why? Is this a rule
> or a bug?
>
> What if I just want to write to the file and nothing else?
>
> Please, look at my tiny program, I wrote this to test mmap(), it is cute.
>
> #include <stdio.h>
>
> typedef struct{
> char name[4];
> int age;
> } people;
>
> int main(int argc, char** argv) {
> int fd, i;
> people *p_map;
> char temp;
>
> fd = open(argv[1], O_RDWR|O_TRUNC); //this works fine
>
> //fd = open(argv[1], O_WRONLY|O_TRUNC); this will cause segmentation fault
>
> if (fd == -1) {
> perror("open()");
> }
> ls

(fd, sizeof(people)*5-1, SEEK_SET);
> write(fd, " " , 1);
>
> p_map = (people *)mmap(NULL,sizeof(people)*5,PROT_WRITE,
MAP_SHARED,fd,0);
> close(fd);
>
> memset(p_map, '\0', sizeof(people)*5);
>
> temp = 'b';
> for(i=0; i<5; i++) {
> temp++;
> //(p_map+i)->name = "aaa";
> //memcpy((p_map+i)->name, &temp, 2);
> memset((p_map+i)->name, temp, 3);
> (p_map+i)->age = 20+i;
> }
> printf("initialize over\n");
> for(i=0; i<5; i++) {
> printf("%s\t%d\n", (p_map+i)->name, (p_map+i)->age);
> }
> sleep(10);
>
> munmap(p_map, sizeof(people)*5);
> printf( "umap ok\n" );
>
> return 0;
> }

on UNIX/LINUX machine your mmap system call will fail if you open the
file
in O_WRONLY mode and try to map that file with PROT_WRITE set for
MAP_SHARED
mapping. See the man page of mmap.
so you are trying to write at some invalid address, thats why SIGSEGV.

i would advise you that after every system call, you *must* always
check whether your system call was successful or not.

Erik de Castro Lopo

2004-08-24, 9:02 am

Hao Xu wrote:
>
> Hi everyone!
>
> I found that if you want to write to the memory got by mmap(),

Mmap() is not part of ISO standard C, so this question is
off topic here.

> you have
> to get the file descriptor for mmap() in O_RDWR mode. If you got the
> file descriptor in O_WRONLY mode, then writing to the memory got by
> mmap() will lead to segmentation fault. Anyone knows why? Is this a rule
> or a bug?

If you open the file write only and then read from it, I
would expect that it should not work. How it "doesn't
work" is probably unspecified and a segfault is not an
unreasonable version of "doesn't work".

Erik
--
+-----------------------------------------------------------+
Erik de Castro Lopo nospam@mega-nerd.com (Yes it's valid)
+-----------------------------------------------------------+
`If you want a vision of the future, it is a wireless broadband
network feeding requests for foreign money-laundering assistance
into a human temporal lobe, forever. With banner ads.'
-- John M. Ford

CBFalconer

2004-08-24, 4:34 pm

junky_fellow wrote:
> Hao Xu <traverser@vip.sina.com> wrote:
.... snip ...[color=darkred]
>
> on UNIX/LINUX machine your mmap system call will fail if you open
> the file in O_WRONLY mode and try to map that file with PROT_WRITE
> set for MAP_SHARED mapping. See the man page of mmap. so you are
> trying to write at some invalid address, thats why SIGSEGV.
>
> i would advise you that after every system call, you *must* always
> check whether your system call was successful or not.

Please don't answer off-topic questions here other than to
redirect them to an appropriate newsgroup. Especially don't
answer them with more off-topic data, which won't be properly
evaluated because the knowledgeable people aren't here.

--
Chuck F (cbfalconer@yahoo.com) (cbfalconer@worldnet.att.net)
Available for consulting/temporary embedded and systems.
<http://cbfalconer.home.att.net> USE worldnet address!

Erik de Castro Lopo

2004-08-27, 3:56 pm

Erik de Castro Lopo

2004-08-31, 3:55 am

Sponsored Links