Hi Jirka,
As _stext getting removed in
dso__load_kallsyms
--> symbols__fixup_duplicate (here _stext removed)
as we have same address in /proc/kallsyms for _stext, asm_do_IRQ and
__exception_text_start
So, we need to disable symbols__fixup_duplicate in dso__load_kallsyms,
Prabhat
On Tue, Jun 25, 2013 at 2:37 PM, Jiri Olsa <jo...@redhat.com> wrote:
> On Tue, Jun 25, 2013 at 11:24:56AM +0530, Prabhat Kumar Ravi wrote:
>> On 3.4-rc49, I got following failure running `perf test`:
>>
>> / # perf test -v 1
>> 1: vmlinux symtab matches kallsyms:
>> --- start ---
>> dso__find_symbol_by_name ---- end ----
>> vmlinux symtab matches kallsyms: FAILED!
>>
>> Perf test is failing at dsofind_symbol_by_name
>>
>> where
>>
>> kallsyms_map = machine__kernel_map(&kallsyms, type);
>>
>> sym = map__find_symbol_by_name(kallsyms_map, ref_reloc_sym.name, NULL);
>> if (sym == NULL) {
>> pr_debug("dso__find_symbol_by_name ");
>> goto out;
>> }
>>
>> Here sym is search for "_stext" which is NULL here so perf test fails
>> here only.
>>
>> On investigation found that _stext having same address as asm_do_IRQ
>> and exception_text_start,
>>
>> c00081c0 T asm_do_IRQ
>> c00081c0 T _stext
>> c00081c0 T __exception_text_start
>>
>> so being deleted by symbolsfixup_duplicate in
>>
>> if (choose_best_symbol(curr, next) == SYMBOL_A) {
>>
>> rb_erase(&next->rb_node, symbols); --> symbole
>> getting erase here.
>> goto again;
>> } else {
>>
>>
>
> I recently fixed this test for another kind of error rising from
> the /proc/kallsyms being not as precise as vmlinux file. This one
> looks like another one.
>
> acme's tree:
> b8bae2c perf tests: Omit end of the symbol check failure for test 1
>
>> My doubt is, Is we really need this commit??:
>>
>> commit 3f5a42722b9e78a434d5a4ee5e607dc33c69ac80
>> Author: Anton Blanchard <an...@samba.org>
>> Date: Wed Aug 24 16:40:15 2011 +1000
>>
>> perf symbols: /proc/kallsyms does not sort module symbols
>>
>> kallsyms__parse assumes that /proc/kallsyms is sorted and sets the end
>> of the previous symbol to the start of the current one.
>>
>> Unfortunately module symbols are not sorted, eg:
>>
>> ffffffffa0081f30 t e1000_clean_rx_irq [e1000e]
>> ffffffffa00817a0 t e1000_alloc_rx_buffers [e1000e]
>>
>> Some symbols end up with a negative length and others have a length
>> larger than they should. This results in confusing perf output.
>>
>> We already have a function to fixup the end of zero length symbols so
>> use that instead.
>>
>> Or we can search for other string.??
>
> This one looks important, I think we should fix the test again.
> Maybe bisable duplicates removal for test? Keep removed symbols
> list? ;-)
>
> jirka
--
To unsubscribe from this list: send the line "unsubscribe linux-perf-users" in
the body of a message to majord...@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
