mmap returning MAP_FAILED

[Washington Ratso]

I am running Linux 2.6.26 on an MPC8248. In my application, mmap is
returning MAP_FAILED. The arguments to mmap are start = 0, length =
2680, prot = 3, flags = MAP_SHARED, fd = 5, offset = 0. errno = 22
"Invalid argument"

I looked at the man page for mmap and don't see a reason why this is
happening. Any ideas?

[David Schwartz]

It would help if you told us what you were trying to do. What is file
descriptor 5 and how was it opened? Also, I doubt '2680' is a multiple
of your page size, so you cannot create a 2,680-byte mapping.

DS

[Washington Ratso]

I am trying to get oprofile to work. The file I am opening is a
sample file. Here is the code for the open:

data->fd = open(filename, flags, 0644);

where filename = "/var/lib/oprofile/samples//current/{kern}/no-vmlinux/
{dep}/{kern}/no-vmlinux/TIMER.0.0.all.all.all"
where flags = O_CREAT | O_RDWR

The file /var/lib/oprofile/samples//current/{kern}/no-vmlinux/{dep}/
{kern}/no-vmlinux/TIMER.0.0.all.all.all does exist.

I changed the length to the page size and it made no difference.
According to http://www.ecst.csuchico.edu/~beej/guide/ipc/mmap.html,
"This can be any length you want. (Aside: if len not a multiple of the
virtual memory page size, you will get a blocksize that is rounded up
to that size. The extra bytes will be 0, and any changes you make to
them will not modify the file.) "

[David Schwartz]

On Nov 6, 6:10 pm, Washington Ratso <jobhunt...@aol.com> wrote:

> I am trying to get oprofile to work.  The file I am opening is a
> sample file.  Here is the code for the open:
>
> data->fd = open(filename, flags, 0644);
>
> where filename = "/var/lib/oprofile/samples//current/{kern}/no-vmlinux/
> {dep}/{kern}/no-vmlinux/TIMER.0.0.all.all.all"
> where flags = O_CREAT | O_RDWR
>
> The file  /var/lib/oprofile/samples//current/{kern}/no-vmlinux/{dep}/
> {kern}/no-vmlinux/TIMER.0.0.all.all.all does exist.

What exactly do you think mapping a zero-length file would do?

DS

[Shankar]

prot=3 => would mean PROT_READ|PROT_WRITE

MAP_SHARED I think is a invalid mode if you are trying to map it that
way, Probably you may want to remove PROT_WRITE and see whats
happening.

Thanks

Shankar Easwaran

[Joe Pfeiffer]

Shankar <shan...@gmail.com> writes:

That would do terrible things to the utility of shared memory -- no,
MAP_SHARED and PROT_WRITE are compatible. Maybe you're thinking of
MAP_SHARED and MAP_ANONYMOUS (which didn't used to be compatible, but is
since 2.4).
--
As we enjoy great advantages from the inventions of others, we should
be glad of an opportunity to serve others by any invention of ours;
and this we should do freely and generously. (Benjamin Franklin)

[Alan Curry]

In article <1br5s9o...@snowball.wb.pfeifferfamily.net>,

Joe Pfeiffer <pfei...@cs.nmsu.edu> wrote:
>Shankar <shan...@gmail.com> writes:
>
>> On Nov 5, 9:24�pm, Washington Ratso <jobhunt...@aol.com> wrote:
>>> I am running Linux 2.6.26 on an MPC8248. � In my application, mmap is
>>> returning MAP_FAILED. �The arguments to mmap are start = 0, length =
>>> 2680, prot = 3, flags = MAP_SHARED, fd = 5, offset = 0. �errno = 22
>>> "Invalid argument"
>>>
>>> I looked at the man page for mmap and don't see a reason why this is
>>> happening. �Any ideas?
>>
>> prot=3 => would mean PROT_READ|PROT_WRITE
>>
>> MAP_SHARED I think is a invalid mode if you are trying to map it that
>> way, Probably you may want to remove PROT_WRITE and see whats
>> happening.
>
>That would do terrible things to the utility of shared memory -- no,
>MAP_SHARED and PROT_WRITE are compatible. Maybe you're thinking of
>MAP_SHARED and MAP_ANONYMOUS (which didn't used to be compatible, but is
>since 2.4).

Also, MAP_SHARED is the only way to get changes written back into a
memory-mapped file, so sometimes you need MAP_SHARED with PROT_WRITE
even if you don't want shared memory (simultaneously accessible by
multiple processes). It's kind of a misleading name.

--
Alan Curry

[Cliff Brake]

On Nov 6, 9:10 pm, Washington Ratso <jobhunt...@aol.com> wrote:

> I am trying to get oprofile to work.  The file I am opening is a
> sample file.  Here is the code for the open:

Make sure you are not logging data to a JFFS2 file system. JFFS2 does
not support MMAP.

Cliff

[karthikbalaguru]

No. It is feasible.
The parameter prot determines whether read, write, execute,
or some combination of accesses are permitted to the data being
mapped.
The prot should be either PROT_NONE or the bitwise inclusive OR of
one or more of the other flags listed below -
PROT_READ Data can be read.
PROT_WRITE Data can be written.
PROT_EXEC Data can be executed.
PROT_NONE Data cannot be accessed.

Check the below links w.r.t mmap ->
http://docs.blackfin.uclinux.org/doku.php?id=:mmap#mmap_kernel_code_example_2
http://www.opengroup.org/onlinepubs/007908775/xsh/mmap.html

Karthik Balaguru

Karthik Balaguru