From: Paulo da Silva <psdasilva@esoterica.pt>
To: Anton Altaparmakov <aia21@cam.ac.uk>
Cc: linux-kernel@vger.kernel.org
Subject: Re: Accessing file mapped data inside the kernel
Date: Thu, 10 Nov 2005 01:44:22 +0000 [thread overview]
Message-ID: <4372A5F6.7030306@esoterica.pt> (raw)
In-Reply-To: <Pine.LNX.4.64.0511092143400.19282@hermes-1.csi.cam.ac.uk>
Anton Altaparmakov wrote:
>Hi,
>
>On Wed, 9 Nov 2005, Paulo da Silva wrote:
>
>
>>I posted about this a few days ago but got no responses
>>so far! I think this should be a trivial question for those
>>involved in the kernel internals. May be I didn't develop
>>the problem enough to be understood.
>>
>>So, here is the question reformulated.
>>
>>A given file system must supply a procedure for mmap.
>>
>>int <fsname>_file_mmap(struct file * file, struct vm_area_struct * vma)
>>{
>> int addr;
>> addr=generic_file_mmap(file,vma);
>> <Code to access addr pointed bytes or vma->vm_start>
>> return addr;
>>}
>>
>>I could verify that "addr" is what is returned to the user as
>>a pointer to a string of bytes that maps a file when a user
>>program calls mmap or mmap2.
>>
>>In the user program, I can access those bytes (read/write)
>>as, for ex., a char pointer.
>>
>>I don't know how to access those bytes inside the kernel
>>at the point <Code to access addr pointed bytes or vma->vm_start>
>>
>>First trys led the program that invoked mmap to block.
>>I thought that there's something to do with a previous
>> down_write(¤t->mm->mmap_sem);
>>If I execute
>> up_write(¤t->mm->mmap_sem);
>>before accessing the data the block situation does not
>>occur anymore. I would like to hear something about
>>this.
>>
>>Anyway, I tryed to use "copy_from_user" but I got
>>garbage, not the file contents! Using "strncpy" crashes
>>the kernel (UML)!
>>
>>Can someone please write a fragment of code to safely
>>access those bytes, copying them to and from a
>>kernel char pointed area so that they are read/written
>>to the file?
>>
>>
>
>Why do you want to do that? If you explain what you are trying to do it
>may be possible to help you better. It is almost 100% certain that your
>are going about it in completely the wrong way, so please describe what
>you are trying to do...
>
>Best regards,
>
> Anton
>
>
Just try to understand the kernel filesystem.
So far I could understand the 1st layer of
reading and writing. mmap seems to be a
difficult task however. So, I made a 1st try
looking at mmap supplied by the filesystem,
but I couldn't even succeed with a printk
of the mapped area! I would like to understand
what is the meaning of the address (int) returned
by generic_file_mmap that is also into vma->vm_start
and is returned to the user as a char pointer.
I thought that this address, being accessible
by a user program as a char pointer, should also
be accessible by a copy-from-user inside the
kernel. Unfortunately, this didn't happen!
Why? That's my question. Did I make any mistake?
A basic fragment of code showing how to access
that area could enlight me so that I could go
deeply into the code.
Ex.
Suppose a file has a string of text ("foo")
and the user calls mmap.
Why does this code not work?
The supplied filesystem mmap is "generic_file_mmap".
So, I changed it to foo_file_mmap as follows:
int foo_file_mmap(struct file * file, struct vm_area_struct * vma)
{
int addr;
char tstr[100];
addr=generic_file_mmap(file,vma);
up_write(¤t->mm->mmap_sem); /* Without this the user program is dead locked */
copy_from_user(tstr,(char*)addr,4);
printk("%s",tstr);
return addr;
}
next prev parent reply other threads:[~2005-11-10 1:44 UTC|newest]
Thread overview: 6+ messages / expand[flat|nested] mbox.gz Atom feed top
2005-11-09 20:15 Paulo da Silva
2005-11-09 20:46 ` linux-os (Dick Johnson)
2005-11-09 21:44 ` Anton Altaparmakov
2005-11-10 1:44 ` Paulo da Silva [this message]
2005-11-10 9:24 ` Anton Altaparmakov
2005-11-10 23:20 ` Paulo da Silva
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=4372A5F6.7030306@esoterica.pt \
--to=psdasilva@esoterica.pt \
--cc=aia21@cam.ac.uk \
--cc=linux-kernel@vger.kernel.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox
all inboxes | Powered by JetHome®