[Crash-utility] fix_lkcd_address problem

[Dave Anderson]

Alan Tyson wrote:
> Hi,
> 
> I believe that there is an incorrect comparison in fix_lkcd_address:
> 
> 059 ulonglong
> 060 fix_lkcd_address(ulonglong addr)
> 061 {
> 062     int i;
> 063     ulong offset;
> 064
> 065     for (i = 0; i < lkcd->fix_addr_num; i++) {
> 066         if ( (addr >=lkcd->fix_addr[i].task) &&
> 067                 (addr <= lkcd->fix_addr[i].task + STACKSIZE())){
>                         ^^^^^- here
> 
> On Itanium fix_addr[i] + STACKSIZE() may be the address of an adjacent 
> task structure.  As it stands both parts of the comparison pass if addr is
> the address in the fix_addr[i].task field or if it is the task structure 
> which follows that one.  The result is this it is not possible to read the
> task structure of the task that follows a task which is in this fixup list
> and zeroes are returned instead.
> 
> Regards,
> 
> Alan Tyson, HP.

Hi Alan,

I normally defer all LKCD-related fixes to Troy Heber,
but unless he disapproves for whatever reason, this one
looks pretty obvious, and I've queued it for the next release.

Thanks,
   Dave

> 
> --- lkcd_common.c.orig  2007-08-27 16:51:11.000000000 +0100
> +++ lkcd_common.c       2007-09-19 16:46:07.000000000 +0100
> @@ -64,7 +64,7 @@ fix_lkcd_address(ulonglong addr)
> 
>     for (i = 0; i < lkcd->fix_addr_num; i++) {
>        if ( (addr >=lkcd->fix_addr[i].task) &&
> -               (addr <= lkcd->fix_addr[i].task + STACKSIZE())){
> +               (addr < lkcd->fix_addr[i].task + STACKSIZE())){
> 
>            offset = addr - lkcd->fix_addr[i].task;
>            addr = lkcd->fix_addr[i].saddr + offset;
> 
> 
> 
> -- 
> Crash-utility mailing list
> Crash-utility at redhat.com
> https://www.redhat.com/mailman/listinfo/crash-utility