how to debug thread stuck in filemap_fault

[shuang liu]

Hi, there, 

some user space threads, say thread-X stuck in  sleep_on_page_killable @ filemap_fault(). 

my question is : 
in normal flow, how to wakeup thread-X from wait_on_page_locked_killable() ? How can I go further to debug this issue? 
I try to use ftrace function_graph to trace the flow, but it can only log EXPORT_SYMBOL_GPL() functions. I am struggling with huge memory and file system source code... 
hope anyone can give me some practical tips. 

Call Trace:
 [<c1895675>] schedule+0x35/0x50
 [<c1895708>] io_schedule+0x78/0xb0
 [<c12c95fd>] sleep_on_page_killable+0xd/0x40
 [<c1895f3f>] __wait_on_bit+0x4f/0x70
 [<c12cbcae>] wait_on_page_bit_killable+0x8e/0x90
 [<c12cbd42>] __lock_page_or_retry+0x92/0xb0
 [<c12cc041>] filemap_fault+0x2e1/0x400
 [<c12e55be>] __do_fault+0x3e/0x390
 [<c12e72ac>] handle_pte_fault+0x7c/0x930
 [<c12e7bf3>] handle_mm_fault+0x93/0xc0
 [<c189a98e>] do_page_fault+0x10e/0x4b0
 [<c189881b>] error_code+0x5f/0x64

here is the kernel code in filemap.c 
static inline int wait_on_page_locked_killable(struct page *page)
{
if (PageLocked(page))
return wait_on_page_bit_killable(page, PG_locked);
return 0;
}

here is ftrace result: 
 3)               |    filemap_fault() {
 3)               |      find_get_page() {
 3)   0.259 us    |        __rcu_read_lock();
 3)   0.217 us    |        __rcu_read_unlock();
 3)   7.432 us    |      }
 3)   0.244 us    |      __might_sleep();
 3) + 12.206 us   |    }
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.kernelnewbies.org/pipermail/kernelnewbies/attachments/20130515/c8463ebf/attachment-0001.html