From mboxrd@z Thu Jan 1 00:00:00 1970 Message-ID: <3BBDE5AF.1A7D5673@mvista.com> Date: Fri, 05 Oct 2001 12:54:07 -0400 From: Dan Malek MIME-Version: 1.0 To: Steve Rossi Cc: John_McRoberts@mitel.com, "linuxppc-embedded@lists.linuxppc.org" Subject: Re: Hard disk and segmentation violation. References: <85256ADC.00515614.00@kanmta01.software.mitel.com> <3BBDDF94.9728840D@labs.mot.com> Content-Type: text/plain; charset=us-ascii Sender: owner-linuxppc-embedded@lists.linuxppc.org List-Id: Steve Rossi wrote: > ..... Still I don't yet have a good > grasp on what space can be mapped 1:1 ..... You shouldn't be concerned about this mapping. ALL device drivers should perform the ioremap() in their device initialization functions, in which case there isn't any mapping problem. There are some unique cases when porting boards or processors that early mapping is necessary, but this is not the place for mapping the resources used by device drivers that initialize normally. Please do not add any resource mapping in the memory management initialization functions. In summary (which has been written too many times already), the user space is first two or three gigabyte virtual address region (zero to 0x7fffffff or 0xbfffffff). Anything mapped in this area obviously conflicts with user mapping. The region between 0x80000000 and 0xbfffffff may be user or kernel space depending upon the kernel configuration and board requirements. The kernel virtual space is always the top gigabyte. With the exception of some special case initialization mapping I mentioned above, the kernel VM allocator dynamically allocates in the proper address spaces. -- Dan ** Sent via the linuxppc-embedded mail list. See http://lists.linuxppc.org/