Query regarding kernel memory

Vaibhav Jain vjoss197 at gmail.com
Tue Jul 26 00:40:44 EDT 2011


Hi Dave, Santhosh,

Thanks for the reply! I was talking about the following paragraph from the
reference you provided :

The kernel (on the x86 architecture, in the default configuration) splits
the 4-GB virtual address space between user-space and the kernel; the same
set of mappings is used in both contexts. A typical split dedicates 3 GB to
user space, and 1 GB for kernel space.* The kernel’s code and data
structures must fit into that space, but the biggest consumer of kernel
address space is virtual mappings for physical memory. The
kernel cannot directly manipulate memory that is not mapped into the
kernel’s address space. The kernel, in other words, needs its own virtual
address for any memory it must touch directly. *Thus, for many years, the
maximum amount of physical memory that could be handled by the kernel was
the amount that could be mapped into the kernel’s portion of the virtual
address space, minus the space needed for the kernel code itself. As a
result, x86-based Linux systems could work with a maximum of a little under
1 GB of physical memory.*

I am still not clear about the sentences in bold. Why is  the space needed
for kernel code subtracted from the amount that could be mapped
into kernel's portion of virtual address space ? Also , what difference does
the fact - "biggest consumer of kernel address space is virtual mappings for
physical memory" make in the amount of memory that can be handled by the
kernel.
I am little confused.

Thanks
Vaibhav Jain

On Mon, Jul 25, 2011 at 5:51 PM, Dave Hylands <dhylands at gmail.com> wrote:

> Hi Vaibhav,
>
> On Mon, Jul 25, 2011 at 3:17 PM, Vaibhav Jain <vjoss197 at gmail.com> wrote:
> > Hi,
> >
> > I read a few articles on linux virtual memory management such as this one
> :
> > http://lwn.net/Articles/75174/
> >
> > which say that earlier linux kernel could only use memory slightly below
> 1
> > GB. They have
> > given the reason for it but I am unable to understand.They further
> describe
> > the use of High memory and low memory.
> > Could anybody please explain the reason for kernel not being able to use
> the
> > 1 GB completely?
> > Also please provide references for high memory and low memory.
>
> My numbers/comments are for the ARM processor, the x86 may be slightly
> different.
>
> The typical configuration for the kernel has addresses from 0x00000000
> through 0xC0000000 given to user space (it's actually a small amount
> less than 3 GB since modules are loaded in the space just before
> 0xC0000000).
>
> That leaves 0xC0000000 to 0xFFFFFFFF for kernel virtual memory (or 1
> Gb). Now devices need some I/O space, which takes away from the 1Gb.
>
> I think what you're calling low memory is kernel logical memory. See
> http://lwn.net/images/pdf/LDD3/ch15.pdf on page 414 (not the 414th
> page of the PDF, but the page with 414 printed on the bottom).
>
> High memory is memory which is not directly accessible by the kernel.
> You need to use kmap/kunmap to map the memory into the kernel virtual
> memory space. Low memory is always accessible by the kernel.
>
> So user-mode programs get allocated from high-memory (if high memory
> exists) since the kernel doesn't typically need to access the
> user-space memory.
>
> It is possible to set some CONFIG options and have the 3Gb/1Gb split
> changed to 2Gb/2Gb or 1Gb/3Gb, but 3Gb/1Gb is the normal default.
>
> --
> Dave Hylands
> Shuswap, BC, Canada
> http://www.davehylands.com
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.kernelnewbies.org/pipermail/kernelnewbies/attachments/20110725/3732de05/attachment.html 


More information about the Kernelnewbies mailing list