Hi, For x86, crash avoids storing any '__crc_*' symbols. It should do the same for ARM, right? Credit goes to Rabin Vincent for this patch, unless you don't like it, in which case you can blame me. Another thing. The ARM kernel potentially includes a symbol 'PRRR' with a value of 0xff0a81a8, defined in arch/arm/mm/proc-v7-2level.S. The problem with this is that it's the symbol which ends up st->symtable[st->symcnt-1] instead of '_end' which means a lot of values will pass this check in in_ksymbol_range(): if ((value >= st->symtable[0].value) && (value <= st->symtable[st->symcnt-1].value)) { if ((st->flags & PERCPU_SYMS) && (value < st->first_ksymbol)) return FALSE; else return TRUE; } How would you prefer dealing with this? How about excluding any symbols with values > '_end'? A KSYMS_END flag could be added to the machdep->flags. Or just unsetting KSYM_START when '_end' is encountered in verify_symbol(). Regards, Per

Hi, For x86, crash avoids storing any '__crc_*' symbols. It should do the same for ARM, right? Credit goes to Rabin Vincent for this patch, unless you don't like it, in which case you can blame me.

On Tue, Jan 29, 2013 at 3:46 PM, Dave Anderson <anderson(a)redhat.com&gt; wrote: > > > ----- Original Message ----- >> Hi, >> >> For x86, crash avoids storing any '__crc_*' symbols. It should do the >> same for ARM, right? Credit goes to Rabin Vincent for this patch, >> unless you don't like it, in which case you can blame me. > > I suppose. It was more of a clutter-reducer than anything else in > the case of x86. But I do note that the __crc_ symbols come after > "_end" in one of the sample ARM dumpfiles I've got on hand. > Clutter reduction is the reason for both of these patches as well and even if the __crc_'s are ignored we'll still get _end+<offset> printouts for everything in the range [_end, 0xff0a81a8]. Of course, a patch for PRRR shouldn't break 'mod -s'. Here's an alternative patch. /Per

> Anyway, I don't have a problem with the patch. Mika? > > Dave > > -- > Crash-utility mailing list > Crash-utility(a)redhat.com > https://www.redhat.com/mailman/listinfo/crash-utility -- Crash-utility mailing list Crash-utility(a)redhat.com https://www.redhat.com/mailman/listinfo/crash-utility

----- Original Message ----- > On Tue, Jan 29, 2013 at 3:46 PM, Dave Anderson <anderson(a)redhat.com&gt; > wrote: > > > > > > ----- Original Message ----- > >> Hi, > >> > >> For x86, crash avoids storing any '__crc_*' symbols. It should do the > >> same for ARM, right? Credit goes to Rabin Vincent for this patch, > >> unless you don't like it, in which case you can blame me. > > > > I suppose. It was more of a clutter-reducer than anything else in > > the case of x86. But I do note that the __crc_ symbols come after > > "_end" in one of the sample ARM dumpfiles I've got on hand. > > > > Clutter reduction is the reason for both of these patches as well and > even if the __crc_'s are ignored we'll still get _end+<offset> > printouts for everything in the range [_end, 0xff0a81a8]. Of course, a > patch for PRRR shouldn't break 'mod -s'. Here's an alternative patch. > > /Per OK good -- I'll queue both the __crc_ and PRRR/NMRR patches for crash-6.1.3, pending any (unlikely) disapprovals/NAKs from Mika or Jan.