Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > comp.unix.programmer > #495

Re: bzero(mmap(NULL, 12288, ANON), 8008) raising bus error!

Show all headers | View raw

On 05/11/11 02:22 PM, loozadroog wrote:
> On May 10, 8:24 pm, Ian Collins<ian-n...@hotmail.com>  wrote:
>> On 05/11/11 10:24 AM, loozadroog wrote:
>>
>>
>>
>>> I just posted a version of this program last night in clc,
>>> asking for a general look-through. But now I've run into
>>> a problem when using mmap that doesn't happen with
>>> malloc.
>>
>>> I'm getting a bus error in memset, attempting to clear
>>> the new memory. What could be going wrong?
>>
>>> Some relevant variables from gdb at the point of crash:
>>> (gdb) r
>>> Starting program: /home/olpc/xpost3/proto/v
>>
>>> Program received signal SIGBUS, Bus error.
>>> 0xb7ec079f in memset () from /lib/libc.so.6
>>> (gdb) backtrace
>>> #0  0xb7ec079f in memset () from /lib/libc.so.6
>>> #1  0x08048799 in mfalloc (mem=0x8049e90, sz=8008) at v.c:79
>>> #2  0x080487b7 in initmtab (mem=0x8049e90) at v.c:99
>>> #3  0x080489c5 in init () at v.c:162
>>> #4  0x080489f1 in main () at v.c:171
>>> (gdb) up
>>> #1  0x08048799 in mfalloc (mem=0x8049e90, sz=8008) at v.c:79
>>> 79          memset(mem->base + adr, 0, sz); /* bus error with MMAP !
>>> */
>>> (gdb) p adr
>>> $1 = 0
>>> (gdb) p sz
>>> $2 = 8008
>>> (gdb) p pgsz
>>> $3 = 4096
>>> (gdb) p mem->max
>>> $4 = 12288
>>> (gdb) quit
>>
>> It looks like you map pgsz (4096) bytes here:
>>
>> mem->base = mmap(NULL, pgsz, PROT_READ|PROT_WRITE, MAP_SHARED
>>               |MAP_ANONYMOUS,-1,0 );
>>
>> and attempt to write sz (8008) bytes here:
>>
>> memset(mem->base + adr, 0, sz);
>>
>> You have the same error with malloc, you just don't see it because the
>> write does not enter unmapped memory, it just pisses on the heap!
>
> But gdb tells mem->max is 12288, so mfalloc must have called growmem,
> and therefore either mremap or realloc has increased the allocation.

I see, well I couldn't test the code, mremap is Linux specific.

> Wouldn't it be a segv if I had overrun the region?

I would have thought so.  Why don't you try something like valgrind to 
look for memory access problems?

> Bus error seems to suggest a problem with the width of the data being
> accessed.

memset doesn't have any alignment restrictions.

>> By the way, why this line?
>>
>> #define _GNU_SOURCE
>
> To activate mremap.

So I see!

-- 
Ian Collins

Back to comp.unix.programmer | Previous | Next — Previous in thread | Next in thread | Find similar | Unroll thread

Thread

bzero(mmap(NULL, 12288, ANON), 8008) raising bus error! loozadroog <mijoryx@yahoo.com> - 2011-05-10 15:24 -0700
  Re: bzero(mmap(NULL, 12288, ANON), 8008) raising bus error! Ian Collins <ian-news@hotmail.com> - 2011-05-11 13:24 +1200
    Re: bzero(mmap(NULL, 12288, ANON), 8008) raising bus error! loozadroog <mijoryx@yahoo.com> - 2011-05-10 19:22 -0700
      Re: bzero(mmap(NULL, 12288, ANON), 8008) raising bus error! Ian Collins <ian-news@hotmail.com> - 2011-05-11 14:39 +1200
        Re: bzero(mmap(NULL, 12288, ANON), 8008) raising bus error! loozadroog <mijoryx@yahoo.com> - 2011-05-10 23:02 -0700
  Re: bzero(mmap(NULL, 12288, ANON), 8008) raising bus error! pacman@kosh.dhis.org (Alan Curry) - 2011-05-11 06:40 +0000
    Re: bzero(mmap(NULL, 12288, ANON), 8008) raising bus error! loozadroog <mijoryx@yahoo.com> - 2011-05-11 17:38 -0700

csiph-web