EFAULT from file read.

EFAULT from file read.

[David Chow]

Dear all,

I am having trouble in reading a file in the kernel space using the
file->f_op->read call, everything is ok. I start off file->f_pos = 0 . I
also did a mntget to the super block before I call
"file=dentry_open(.....)" . I intend to open the file in read only mode.
What can be wrong? I have also check the inode->i_size is large enough
for me to just read 8 bytes from the file. I keep having EFAULT error
from the read call... also before calling mntget, also did a dget to the
dentry. Any help is welcomed. Thanks.

regards,

David Chow

Re: EFAULT from file read.

[Richard B. Johnson]

On Tue, 18 Sep 2001, David Chow wrote:

> Dear all,
> 
> I am having trouble in reading a file in the kernel space using the
> file->f_op->read call, everything is ok. I start off file->f_pos = 0 . I
> also did a mntget to the super block before I call
> "file=dentry_open(.....)" . I intend to open the file in read only mode.
> What can be wrong? I have also check the inode->i_size is large enough
> for me to just read 8 bytes from the file. I keep having EFAULT error
> from the read call... also before calling mntget, also did a dget to the
> dentry. Any help is welcomed. Thanks.
> 
> regards,
> 
> David Chow
> -

Are you attempting file access in the kernel from a kernel-thread,
or have you just decided to perform file access from some random
kernel code?

File I/O requires a process context. Your file descriptor means
nothing unless associated with the process that opened the file.
The kernel, itself, is not a process. It functions in the process
context of the caller. Also file I/O sleeps and pages. You need to
understand this before you attempt to access files from within the
kernel.

Normally, you don't access files from within the kernel. If you
need file data for a driver for instance, you write an appropriate
ioctl() function in your driver so a user-mode task can send the
data to the driver. Your startup script inserts the module, then
executes a program which initializes the module with the appropriate
file data and initialization commands, all sent from user mode.
Once the driver is initialized, the user-mode "starter" exits,
freeing resources which would otherwise be trapped forever in your
driver.

Cheers,
Dick Johnson

Penguin : Linux version 2.4.1 on an i686 machine (799.53 BogoMips).

    I was going to compile a list of innovations that could be
    attributed to Microsoft. Once I realized that Ctrl-Alt-Del
    was handled in the BIOS, I found that there aren't any.

Re: EFAULT from file read.

[Alexander Viro]

On Tue, 18 Sep 2001, Richard B. Johnson wrote:

> File I/O requires a process context. Your file descriptor means
> nothing unless associated with the process that opened the file.

It fscking doesn't.  He had clearly said that he calls file->f_op->read(),
which has nothing whatsofuckingever to descriptors.  Sod off and don't
return until you learn to read.

As for the original question - grep fro set_fs and you'll see what to
do (basically, set_fs(KERNEL_DS) before the call of ->read() and restore
afterwards).

Re: EFAULT from file read.

[David Chow]

Alexander Viro ¼g¹D¡G
> 
> On Tue, 18 Sep 2001, Richard B. Johnson wrote:
> 
> > File I/O requires a process context. Your file descriptor means
> > nothing unless associated with the process that opened the file.
> 
> It fscking doesn't.  He had clearly said that he calls file->f_op->read(),
> which has nothing whatsofuckingever to descriptors.  Sod off and don't
> return until you learn to read.
> 
> As for the original question - grep fro set_fs and you'll see what to
> do (basically, set_fs(KERNEL_DS) before the call of ->read() and restore
> afterwards).

Problem solved by calling dummy=set_fs(KERNEL_DS) . But remember to call
set_fs(dummy) to restore after the read.

regards,

David