[Info-vax] The (now lost) future of Alpha.

[Johnny Billquist]

On 2018-08-08 19:57, Bob Koehler wrote:
> In article <pkcs69$btk$1 at Iltempo.Update.UU.SE>, Johnny Billquist <bqt at softjar.se> writes:
>>
>> Yes. When we say "ioctl" we can talk about two different things. Any
>> user mode program who calls ioctl() is calling a routine in user space,
>> which is included from a library. This routine is, by necessity, written
>> in assembler. But it is running in user space.
>>
>> This routine in the library sets up the sequence to do the actual system
>> call.
> 
>     Nope.  The user mode routine was built into the kernel, and of
>     course, the kernel mode routine was also.  Not a separate library.

Please point to that code. Which OS? And I would hope the code is 
available, so you can give a link.

How do you think the transfer from the users program to the kernel user 
mode code happen? Do you then treat the kernel as a dynamic library, for 
which addresses are resolved at runtime, and you do a direct jump into 
the kernel? Or do you resolve addresses at link time, and the code 
becomes dependent on the kernel version?

If we're talking about user mode code in the kernel, then it must be 
done through a jump/call and not some trap. And then the first question 
becomes - how do you get the correct address to call to. And if that 
memory is accessible from the rest of your program, you then need to 
separate bits of the kernel to only make some parts available to the 
user process, while other parts still need to be protected from random 
access by the user program.

   Johnny

-- 
Johnny Billquist                  || "I'm on a bus
                                   ||  on a psychedelic trip
email: bqt at softjar.se             ||  Reading murder books
pdp is alive!                     ||  tryin' to stay hip" - B. Idol