NetBSD-Bugs archive

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Old Index]

Re: port-powerpc/39257: Mistake in sys/arch/powerpc/powerpc/trap_subr.S file

Havard Eidnes pisze:
Tim Rightnour pisze:
 I looked at the file at the location you mentioned, but I don't
 think the current code in CVS matches whatever you are looking
 at.  The line numbers you gave don't match up.  Can you please
 either tell me what the CVS revision number at the top of the
 file you are looking at is, or look at the most current one,
 which is
Dear Tim,
Thank you very much for quick answer. You are right, I was looking at trap_subr.S v1.59.
In v 1.64 the line numbers are 540-547.
Lines 540 and 541: in my opinion the contents of R1 are not used anymore.
Line 544: change of the contents of R2 results in incorrect value loaded into R1 in line 547.

It would probably be helpful if you quoted the actual code and gave
a bit more context, such as which function you are talking about.
The reason I say this is that in revision 1.64 of trap_subr.S, lines
540-547 look as follows:

_C_LABEL(tlbdlmsize) = .-_C_LABEL(tlbdlmiss)

/* LINTSTUB: Var: int tlbdsmiss[1], tlbdsmsize[1]; */
        .globl  _C_LABEL(tlbdsmiss),_C_LABEL(tlbdsmsize)
        mfspr   %r2,SPR_HASH1           /* get first pointer */
        li      %r1,%r8

which is probably not what you wanted to indicate.


- Havard
Dear Havard,

Sorry again. German NetBSD FTP server has out of date NetBSD-current.
The lines are 590-597:

   mfspr    %r3,SPR_DMISS        /* get the miss address */
   mfsrin    %r1,%r3            /* get the segment register */
   mfsrr1    %r3
   rlwinm    %r3,%r3,18,31,31    /* get PR-bit */
   rlwnm.    %r2,%r2,%r3,1,1        /* get the key */
   bne-    9b            /* protection violation */
8:    /* found, set reference/change bits */
   ldreg    %r1,4(%r2)        /* reload tlb entry */

At the moment I know which line has a bug. It is line
rlwnm.    %r2,%r2,%r3,1,1        /* get the key */
and it should be:
rlwnm.    %r1,%r1,%r3,1,1        /* get the key */

Now it seems to work OK. How is it possible that you haven't noticed it before?

Kind regards,
Konrad Karpowicz

Home | Main Index | Thread Index | Old Index