fseek error

[Amith_2]

Hello guys,

I have a problem with fseek here.It gives me an error 27 (EFBIG File too large). The file on which i am doing the fseek has reached 1.99GB by size at that time.
Is there any LARGE FILE restriction for fseek. If so how can i resolve this.

Regards
Amith

[Steven E. Protter]

Shalom Amith,

The filesystem was created without largefiles enabled.

Two choices.

1) use newfs with the option -o largefiles to recreate the filesystem.
2) use fsadm to convert the filesystem for large files.

SEP

[Denver Osborn]

ulimit could be a factor (ulimit -f)

Is the file you're working on currently larger than 2GB? If so, then the filesystem where it resides is probably not the issue.

-denver

[Amith_2]

Steve ,

There are some other files in the same file system whcih is of size greater than 4GB. If it is a file system problem then , such files will not be there.

Is there any restriction in fseek ??

Many Thanks
Amith

[Denver Osborn]

ok... looks like you need fseeko(), I saw it in a largfiles whitepaper then searched google for "fseek() fseeko() site:itrc.hp.com" to find this thread.

http://forums1.itrc.hp.com/service/forums/questionanswer.do?threadId=857114&admit=-682735245+1152124371842+28353475

hope it helps,
-denver

[A. Clay Stephenson]

Note that the offset argument is a signed long int; if you compile with _LARGEFILE_SOURCE defined then fseeko which uses an off_t as the offset argument will be used.

One approach that would make your code more portable and POSIX compliant would be to use lseek(), read(), and write() using file descriptors rather than the higher level *FILE abstractions although, you would also need to handle the buffering yourself as well if needed.

[Amith_2]

Hello ,

I tried changing the call to fseeko and compiled using the _LARGEFILE_SOURCE option , still the error persists.
I checkmed the file system attributes in /etc/fstab in that also "largefiles" option is enabled. It is a vxfs file system.

Any inputs ??

[A. Clay Stephenson]

It is actually possible that since you are feeding it a signed long that the value is negative due to underflow; it is also possible that you have hit the processes ulimit since either condition could set errno = EFBIG so if this fails with fseeko using an off_t as the actual parameter then I would suspect ulimit. Just before this fseeko, insert a ulimit(UL_GETFSIZE) and determine the current number of blocks allowed.

[Amith_2]

Hi,

I inserted a ulimit(UL_GETFSIZE) in my program and it displayed 4194303( 4194303 * 512 = 2GB).
But when i type the command from the user prompt , it displays the ulimit as unlimited. The ulimit -a output is
time(seconds) unlimited
file(blocks) unlimited
data(kbytes) 488280
stack(kbytes) 58592
memory(kbytes) unlimited
coredump(blocks) 4194303
nofiles(descriptors) 1024

Could you please tell me when this can happen and how can i resolve it .

Thanks
Amith

[A. Clay Stephenson]

You have now found your fseek() problem. I assume that you executed the ulimit command just before executing the program with the fseek() in it --- and as the same user. Even in that case there is still a simple explanation as to why ulimit is now smaller: there is a ulimit(UL_SETFSIZE,newsize) somewhere in your program -- although it may be in a library object. Note that the most recent ulimit() always wins.