[OpenAFS-devel] Final Mac OS Steps

[Chas Williams (CONTRACTOR)]

do a 'make test' inside src/lwp.  if you can get these sample programs
to work, then the rest should work.

LWP_SP is the offset into the jmp_buf that represents the stack pointer.
jmp_buf is typically a series of longs representing a process context.
lwp bangs the sp inside a jmp_buf in an attempt to manipulate the stack
since isnt a well defined api for this (there is ucontext, but not
everyone implements this).

mac os typically uses the assembly versions of the lwp stack manipulation
code, as i recall.  you might have better luck just figuring out the
arm32 specific assembly you would need to accomplish this task.

In message <49D16F0B.5030000@isscp.com>,Barry Fawthrop writes:
>Once again you are right
>I played around changing the LWP_SP value  in process.c  I got further when set to 9
>
>Now pr_Initialize completes
>and it stops in SNameToId  at the
>code = ubik_PR_NameToID(pruclient, 0, &lnames, &lids);
>
>rxi_ReadPacket calls
>nbytes = rxi_recvmsg(socket, &msg, 0);
>which stops on the
>   return recvmsg(socket, msg_p, flags);
>
>   socket = 6
>   msg_p  = 0x79be80  namelen = 16
>   flags  = 0
>
>
>Previous frame inner to this frame (gdb could not unwind past this frame)
>Previous frame inner to this frame (gdb could not unwind past this frame)
>
>Is there some  place I can learn more about LWP  and or especially
>AFSs usage of LWP?
>
>The comments give some insight but not enough to grasp and understand what is going on ..
>
>
>Thanks again
>
>
>
>
>Derrick Brashear wrote:
>> On Mon, Mar 30, 2009 at 1:31 PM, Barry Fawthrop <barry@isscp.com> wrote:
>> 
>>> Yes Derrick, you are quite right.
>>>
>>> This does lead me to two questions
>>>
>>> (1) What is the need for the kernel module portion and what are it's
>>> functions
>>> could this be managed some other way, perhaps without a kernel module?
>>>
>> 
>> it's to make files available as files. you could do it in an application but
>> you'd need to do it in every application which you wanted to have AFS
>> access. not much fun there.
>> 
>>> (I'm no expert in these things, but I do have a good grasp of computing
>>> and always willing to learn more)
>>>
>>>
>>> (2) I'm finding that pr_Initialize fails when it get to
>>> process.c  savecontext
>>>
>> 
>> LWP_SP et al are probably wrong for your hardware.
>> 
>
>_______________________________________________
>OpenAFS-devel mailing list
>OpenAFS-devel@openafs.org
>https://lists.openafs.org/mailman/listinfo/openafs-devel
>