Fiasco.OC: null-pointer dereference?

Leslie Zhai lesliezhai at llvm.org.cn
Fri Jun 9 04:42:47 CEST 2017


Hi Jean,

Thanks for your help!


在 2017年06月09日 00:53, Jean Wolter 写道:
> On 08/06/17 04:18, Leslie Zhai wrote:
>> Hi Matthias,
>>
>> Jean taught me about how to debug L4Re using jdb in qemu 
>> http://os.inf.tu-dresden.de/pipermail/l4-hackers/2017/008038.html it 
>> used a on purpose bug (null ptr deref) to crash Ned, then L4Re 
>> thrown: unhandled write page fault at 0x0 pc=0x100398d, and addr2line 
>> ... -e ned -a 100398d to indicate the root cause line.
>>
>> But how to find out the root cause if unclear that which components 
>> bring in the issue?
>
> I think there might be a misunderstanding. I only introduced the null 
> pointer dereference to demonstrate how to do it using a known problem. 
> You can apply exactly the same steps in a different situation.
I just want to express the ('on purpose' is misuse, sorry for my poor 
English) debug patch is a demo to guide me how to debug with jdb in the 
qemu :) you are my mentor teaching me patiently and carefully!


>
> But I would like to add something. You actually had all the 
> information you needed:
>
>          MOE: loading 'rom/ned'
>          Ned says: Hi World!
>     [1] 0 pf:  0022 pfa=0000000000000018 ip=fffffffff0031ea9 (R-)
>     spc=0xffffffff807c3dd
>     [2] L4Re[rm]: unhandled read page fault at 0x18 pc=0x102e893
>     [3] L4Re: unhandled exception: pc=0xfffffffff0031ea9 (pfa=18)
>          L4Re: Global::l4re_aux->ldr_flags=0
>
>
> In [2] you see the message from the local pager, that is unable to 
> find a valid region for the pagefault address and complains. It shows 
> the 0x18 as pagefault address and an instruction pointer 0x102e893. 
> The instruction pointer did not make any sense at that time. The local 
> pager triggers an exception.
>
> In [3] you see the exception message. It shows the instruction pointer 
> where the pagefault was actually raised: 0xfffffffff0031ea9. This is 
> an address inside the kernel:
That is the key point! it is magic to me that 0xfffffffff0031ea9 is an 
address inside the kernel, I need to deepinto Fiasco about address 
space, correct?


>     ~/build/tmp/l4re$ addr2line -p -i -e
>     ../leslie/fiasco/build/fiasco.image -a fffffffff0031ea9
>     0xfffffffff0031ea9:
>     /home/zhaixiang/project/l4re/kernel/fiasco/src/drivers/amd64/processor-amd64.cpp:67
>     ...
>      (inlined by)
>     /home/zhaixiang/project/l4re/kernel/fiasco/src/kern/ram_quota.cpp:53
>
>     /home/zhaixiang/project/l4re/kernel/fiasco/src/drivers/amd64/processor-amd64.cpp:67
>     fffffffff0031ea8:       fa                      cli
>     fffffffff0031ea9:       48 8b 47 18             mov 0x18(%rdi),%rax
>     fffffffff0031ead:       48 03 77 10             add 0x10(%rdi),%rsi
>     _ZN9Ram_quota5allocEl():
>     /home/zhaixiang/project/l4re/kernel/fiasco/src/kern/ram_quota.cpp:54
>
> If it is not a kernel fault and you need to find out, which component 
> is responsible (or need more information about the current state) you 
> can press 'i' when line [0] appears. You enter the kernel debugger and 
> can look at the current thread using t<enter>. The thread has an id, 
> which you can lookup in the list of present threads (using 'lp'). Here 
> it is thread 22:
>
>       id  cpu    name             pr     sp  wait    to stack state
>        2e   0     -----             2     1e     -       ( 920) rcv_wait
>        2b   0     -----            10     1e     -       (1072) rcv_wait
>        22   0     -----             2     1e             (1776) ready
>        1f   0     #ned             ff     1e     -       (1072)
>     ready,rcv_wait
>
> All threads shown here have the same address space and therefore the 
> problem happened in the context of ned.
>
>>
>> The same story is how to debug L4Linux? 
>> http://os.inf.tu-dresden.de/pipermail/l4-hackers/2017/008047.html 
>> please give me some advice, thanks a lot!
>>
> Maybe you can add "-serial stdio" to your qemu options and provide the 
> complete backtrace for the problem? It looks like a framebuffer issue, 
> but there should be more information in the lines above ...
I will try instead of posting screenshots on Twitter, sorry for my posting!


>
> regards,
> Jean
>

-- 
Regards,
Leslie Zhai - a LLVM hacker https://reviews.llvm.org/p/xiangzhai/

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://os.inf.tu-dresden.de/pipermail/l4-hackers/attachments/20170609/65460fa6/attachment.html>


More information about the l4-hackers mailing list