RE: MMap offset parameter failing with ENOMEM error

["Shane Mann" <shanem at citr dot com dot au>]

Corinna,

I noticed this while I continued testing last night. Interestingly Linux and
SunOS don't seem to be bothered by this, however, if you read the IEEE
standard: http://www.opengroup.org/onlinepubs/007904975/functions/mmap.html
there is the following quote: "The range of bytes starting at off and
continuing for len bytes shall be legitimate for the possible (not
necessarily current) offsets in the file, shared memory object, or [TYM]
typed memory object  represented by fildes." Linux and Sun seem to allow it
and return '0' if you access memory beyond the EOF, go figure.

Perhaps the wrong error is being returned for this case - I would have
thought the following error code would be more appropriate:

[ENXIO]
Addresses in the range [off,off+len) are invalid for the object specified by
fildes.

I have since amended the code by replacing the following line:

size = finfo.st_size;

with

size = finfo.st_size - foff;

and I still get the same error. There is some example output below:

Note: index.idx is a file with a size of 557592 bytes. In the output below
it worked for an offset of 400000 and created a map of 157592 bytes - but
failed for the same file with an offset of 300000 and a potential map of
257592 bytes.

Run 1)
~/c>./a.exe 400000 index.idx
Offset is 400000
File index.idx opened on descriptor 3.  It is 157592 bytes long
--- Syspagesize 4096, pagemask FFFFF000, offmask FFF
foff is 400000
--- Syspagesize 4096, pagemask FFFFF000, offmask FFF
pfoff is 397312
**** Vars just before mmap call ****
psyz = 163840
prot = 1
fd = 3
pfoff = 397312
   Mapping 163840 at 2A231000 <---> 397312 (mapping fh
Result from femmap is 2A231A80
Byte is 127


Run 2)
~/c>./a.exe 300000 index.idx
Offset is 300000
File index.idx opened on descriptor 3.  It is 257592 bytes long
--- Syspagesize 4096, pagemask FFFFF000, offmask FFF
foff is 300000
--- Syspagesize 4096, pagemask FFFFF000, offmask FFF
pfoff is 299008
**** Vars just before mmap call ****
psyz = 262144
prot = 1
fd = 3
pfoff = 299008
perror reports errno of 12
   Mapping 262144 at FFFFFFFF <---> 299008 (mapping fh

Any ideas?

Cheers,
Shane

Shane Mann
Software Engineer          Phone: +61-7-3259-2223
LeadUp Software Pty Ltd    Fax:   +61-7-3259-2259
339 Coronation Drive,      Email: shane.mann@leadup.com.au
Milton, QLD, 4064          Web:   http://www.leadup.com.au


> -----Original Message-----
> From: cygwin-owner@cygwin.com [mailto:cygwin-owner@cygwin.com]On Behalf
> Of Corinna Vinschen
> Sent: Thursday, 19 September 2002 17:33
> To: Cygwin@Cygwin. Com
> Subject: Re: MMap offset parameter failing with ENOMEM error
>
>
> On Thu, Sep 19, 2002 at 12:19:06PM +1000, Shane Mann wrote:
> > Hi,
> >
> > I am having trouble with mmap using the last 'offset' parameter. I have
> > attached a small piece of code (and sample file) which demonstrates the
> > problem. Basically if the offset parameter is passed as a
> variable to mmap
> > then the call fails with an ENOMEM error. But if you pass 0 as
> the parameter
> > and then access the return address from mmap + offset - it is fine. My
> > question: is mmap handling the offset parameter correctly?
> >
> > Usage for the program:
> >
> > ./a.exe <offset> <file> eg: ./a.exe 11200 index.idx
> >
> > To see the address + offset work, comment out the first mmap
> call and return
> > statement and uncomment the lines below each one.
> >
> > Any help appreciated.
>
> Your application tries to mmap over EOF.
>
> Corinna
>
> --
> Corinna Vinschen                  Please, send mails regarding Cygwin to
> Cygwin Developer                                mailto:cygwin@cygwin.com
> Red Hat, Inc.
>
> --
> Unsubscribe info:      http://cygwin.com/ml/#unsubscribe-simple
> Bug reporting:         http://cygwin.com/bugs.html
> Documentation:         http://cygwin.com/docs.html
> FAQ:                   http://cygwin.com/faq/
>


--
Unsubscribe info:      http://cygwin.com/ml/#unsubscribe-simple
Bug reporting:         http://cygwin.com/bugs.html
Documentation:         http://cygwin.com/docs.html
FAQ:                   http://cygwin.com/faq/