Re: [PATCH] perf, x86: Disable sanity check

[Arun Sharma]

On 4/18/12 10:22 PM, Frederic Weisbecker wrote:

> So rbp is part of the JIT stack but not rsp?
> Do you have a practical example of that? I must confess I don't know
> much about JIT stack.

Nothing specific to JITs here. Any time an app has two stacks S1 and S2 
(with S1 at a lower address and S2 at a higher address) and %rsp at the 
time of a perf event is pointing to S2, we don't get traces beyond S2.

              |                 |
              |                 |
              |   0x1000000     |  <-- %rbp
              |                 |
              |                 |
S2: 0x2000000 |                 |  <-- %rsp
              |                 |
              |                 |
              |                 |
              |    frame3       |
              |                 |
              |    frame2       |
              |                 |
S1: 0x1000000 |    frame1       |
              |                 |
              |                 |
              |                 |
              |                 |

gdb is able to show:

(gdb) bt
<frame at 0x2000000>
frame1
frame2
frame3
frame4
..

just fine.

In our use case, there may be multiple transitions between S1 and S2. 
For eg: frame4 could be on S2 (i.e 0x2xxxxxx range).

 -Arun
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/