Mapping perf event files twice

[Soeren Sandmann]

Hi,

In the branch where perf_mmap is backed with vmalloc memory, this
code:

        data->user_page = all_buf;
        data->data_pages[0] = all_buf + PAGE_SIZE;
        data->data_order = ilog2(nr_pages);
        data->nr_pages = 1;

sets data->nr_pages to 1. This interferes with Sysprof which tries to
map the file twice consecutively to handle overflows automatically,
and therefore hits this in perf_mmap():

        if (atomic_inc_not_zero(&event->mmap_count)) {
                if (nr_pages != event->data->nr_pages)
                        ret = -EINVAL;
                goto unlock;
        }

Because event->data->nr_pages is 1 and Sysprof uses 32 pages, it gets
-EINVAL.

Possibly Sysprof shouldn't be playing this game, but if mapping the
file twice is not supported, then maybe perf_mmap() should just
explicitly forbid it.


Thanks,
Soren
--
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/