On Thu 27-Oct-2016 1:12 PM, Jeff Waugh wrote:
On Thu, Oct 27, 2016 at 12:08 PM,
<[email protected]<mailto:[email protected]>> wrote:
Now here's the problem. We want message register two. But on ARM, the first 4
message registers are done actually in register (not the IPC buffer). When we
called seL4_CallWithMRs, it gave returned us message register 2 in the variable
we passed to it, but we have then thrown it away, thus forever losing the error
information.
I can see the more complex generated functions (that use more than the number
of available real registers) are comfy using SetMR before and GetMR after
*WithMRs calls. I'm assuming a SetMR is safe after the *WithMRs call? Correct
me if I've misunderstood something in there, but it seems about right.
Naively, would the following be appropriate to generate:
LIBSEL4_INLINE long
seL4_ARM_Page_Map(seL4_ARM_Page service, seL4_ARM_PageDirectory pd, seL4_Word
vaddr, seL4_CapRights rights, seL4_ARM_VM
Attributes attr)
{
...
/* Perform the call, passing in-register arguments directly. */
output_tag = seL4_CallWithMRs(service, tag,
&mr0, &mr1, &mr2, seL4_Null);
/* Unmarshal real register results into the IPC buffer. */
seL4_SetMR(0, mr0);
seL4_SetMR(1, mr1);
seL4_SetMR(2, mr2);
// 4th register isn't relevant for this function, but would be elsewhere
return seL4_MessageInfo_get_label(output_tag);
}
This is where things start to get complicated. Different errors have a
different number of words for their fault. In the worst case this can be 5
words, so all 4 message registers on arm need to be stored. Without some
detailed information to the syscall stub generator of exactly which invocations
can cause which errors, you're looking at just always stashing all 4 MRs back
into the IPC buffer. Could probably optimize this by checking the label before
return and storing the MRs only if an error happened.
I actually cannot think of a compelling reason for why that wouldn't work....
Either there is something I've forgotten, or the solution really is that simple.
Adrian
_______________________________________________
Devel mailing list
[email protected]
https://sel4.systems/lists/listinfo/devel
