On Wed, Feb 09, 2005 at 08:19:30AM -0800, Srdjan Sobajic hit keys to express the following:
> Hi,
>
> In the 2.4 kernel series, is the following code correct?
>
> copy_to_user(user_buffer, kernel_buffer_a, kernel_buf_a_size);
> copy_to_user(user_buffer + kernel_buf_a_size, buffer_b, buf_b_size);
>
> The point is calling copy_to_user twice to copy two kernel mode blobs
> into the same user mode buffer, while _incrementing_ the user mode
> buffer pointer.
This is correct as far as the kernel is concerned. Before doing
copy_{to,from}_user, kernel will check the access permissions (access_ok() ).
But the functionality of access_ok() is limited to checking whether the address
has reasonable permissions (i.e, the page belonging to memory is legitimately
allocated to the process).
say the code,
char *str = malloc(10);
succeeds. Internally glibc may allocate 1 page for the process as a result of
this request. if you pass str to kernel and try
kbuffer = kmalloc (100, GFP_KERNEL);
copy_from_user ( str, kbuffer, 100); /* this call will succeed, but produce
unsavoury results).
Hope this helps.
Om.
--
Excellent day for putting Slinkies on an escalator.
--
Kernelnewbies: Help each other learn about the Linux kernel.
Archive: http://mail.nl.linux.org/kernelnewbies/
FAQ: http://kernelnewbies.org/faq/
