Ok, I have just fixed it.
The problem was _not_ in the kernel (shame on me that I doubted Linus for
even a microsecond!).
The bug is in the 3dfx_driver.c - in mmap_3dfx() function comment out the
f_count++:
The code at the end should look like this:
#ifdef KERNEL_VER_2_1
vma->vm_file = file;
/* file->f_count++; */
#else
vma->vm_inode = inode;
inode->i_count++;
#endif
Until now it was ok because it was driver's responsibility to increment
f_count, but now do_mmap() does it.
Regards,
Tigran.
On Thu, 11 Mar 1999, Tigran Aivazian wrote:
> Hi,
>
> Yes, it is broken. There were two changes to mm/mmap.c:
>
> a) in do_mmap()
> vma->vm_file = file;
> file->f_count++;
>
> b) in do_munmap()
>
> errno = -EINVAL; instead of 0, because Single UNIX v2 says so.
>
> I have just checked that it is *not* b) that breaks it (thank heaven - for
> b) was my patch :)
>
> It appears that a /dev/3dfx usage count (file->f_count) is incremented but
> not decremented. So I am suspicious of a) - perhaps there should be a
> corresponding file->f_count-- in do_munmap() ?
>
> I am trying to fix it now - will let you know if any luck.
>
> Regards,
> Tigran.
>
>
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/