Re: Value too large for defined data type error text should be: 'Kernel buffer overflow?'

Jens.Toerring_at_physik.fu-berlin.de
Date: 01/29/05 

Date: 29 Jan 2005 18:04:45 GMT

Jan Panteltje <pNaonStpealmtje@yahoo.com> wrote:
> On a sunny day (29 Jan 2005 15:35:40 GMT) it happened
> Jens.Toerring@physik.fu-berlin.de wrote in <361oqcF4p6m0cU1@uni-berlin.de>:

>>Jan Panteltje <pNaonStpealmtje@yahoo.com> wrote:
>>> No teven sure I am asking in the right Usenet group.
>>> I am using DVB-S (satellite) and receiving whol transponder streams,
>>> abiou 10GB / hour or more.
>>> One way is to use 'cat' from the DVB device:
>>> cat /dev/db/adapter0/dvr0 > myfile.ts
>>
>>> When doing this it aborts often (after few minutes) wit hteh message:
>>> 'Value too large for defined data type'
>>> Now tha tfist set me lookin gin the wrong direction.
>>> So I looked it up in:
>>> /usr/src/linux-2.4.25/include/asm-i386/errno.h (pfff took a file to find,
>>> there are lots of errno.h)
>>> It says:
>>> #define EOVERFLOW 75 /* Value too large for defined data type */
>>
>>I guess the problem is related to the file getting too large. When
>>you stream data at a rate of 10 GB/hour into a file you reach the
>>traditional limit of 2 GB
> No that has nothing to do with it, that is exactly WHY that error message
> SHOULD be 'buffer overflow in kernel'.

How do you know it's a "buffer overflow in kernel"? All you told is
that you get an error message that seems to be related to EOVERFLOW.
And if it would really be a buffer overflow in the kernel it's rather
unlikely that you get a nice error message, it's much more likely that
the system crashes...

> I save 8.5GB files with xdipo no problem! (on reiserfs).
> All makefiles have -D_FILE_OFFSET_BITS=64 in them!
> This replaces the 32 bit interface with the 64 bit interface completely
> (see libc.info).
> I did some googling and LOTS of people think that messge relates to files > 2 GB IT DOES NOT.
> It is an overflow error!!!!!!!!!!, that is exactly why I would like to see
> the errno.h text changed!

Now you lost me completely. Why should the text be changed? It's an
overflow of the data type supposed to hold a value - the word "over-
flow" has been used for that kind of things since long before I
started using computers and that's already quite some time, just try
how many hits you get when you search for e.g. "integer overflow".
And changing anything in errno.h wouldn't change the text you get for
the error message (where perror() etc. gets the texts for certain
values of errno from is rather likely somewhere in libc). The text
is already quite ok - don't try to read too much meaning into the
symbolic names for the values errno can assume.

                                 Regards, Jens 

-- 
  \   Jens Thoms Toerring  ___  Jens.Toerring@physik.fu-berlin.de
   \__________________________  http://www.toerring.de

Relevant Pages

  • Re: Value too large for defined data type error text should be: Kernel buffer overflow?
    ... >> SHOULD be 'buffer overflow in kernel'. ... >that you get an error message that seems to be related to EOVERFLOW. ... >And if it would really be a buffer overflow in the kernel it's rather ...
    (comp.os.linux.development.apps)
  • Help with big formula: buffer overflow in read input!
    ... I'd like to plot a function in 3D given by a big formula, ... this error message. ... buffer overflow in read input! ... Prev by Date: ...
    (comp.graphics.apps.gnuplot)
  • Re: Help with big formula: buffer overflow in read input!
    ... > I'd like to plot a function in 3D given by a big formula, ... > buffer overflow in read input! ... version, and the complete, *actual* error message you got. ... Prev by Date: ...
    (comp.graphics.apps.gnuplot)
  • (no subject)
    ... The Kernel returns a ... when I read the value of errno it contains ... error message. ... send the line "unsubscribe linux-kernel" in ...
    (Linux-Kernel)
  • Re: real security - no foreign binaries
    ... I was thinking --- it wouldn't be too hard to make buffer overflow ... attacks impossible. ... I was thinking that part of building a kernel would ... The real security problems are s/w bugs that lead to root> compromise. ...
    (comp.os.linux.security)
Loading