Re: [Qemu-devel] [Bug 824650] [NEW] Latest GIT assert error in arp_table.c

[Jan Kiszka <jan.kiszka@siemens.com>]

On 2011-10-27 05:21, Roy Tam wrote:
> 2011/10/27 Roy Tam <roytam@gmail.com>:
>> Hi,
>>
>> 2011/10/26 Jan Kiszka <jan.kiszka@siemens.com>:
>>> On 2011-10-26 10:03, Roy Tam wrote:
>>>> Hi,
>>>>
>>>> 2011/9/15 Jan Kiszka <jan.kiszka@web.de>:
>>>>> On 2011-09-15 15:20, Roy Tam wrote:
>>>>>> 2011/9/15 Jan Kiszka <jan.kiszka@web.de>:
>>>>>>> On 2011-09-15 14:05, Roy Tam wrote:
>>>>>>>> Here you go.
>>>>>>>>
>>>>>>>> sb16: warning: command 0xf,1 is not truly understood yet
>>>>>>>> sb16: warning: command 0xe,2 is not truly understood yet
>>>>>>>> [Switching to Thread 13840.0x3140]
>>>>>>>>
>>>>>>>> Breakpoint 1, arp_table_search (slirp=0x19f7380, ip_addr=4294967295,
>>>>>>>>     out_ethaddr=0x20af64a "") at slirp/arp_table.c:75
>>>>>>>> 75      //    assert((ip_addr & htonl(~(0xf << 28))) != 0);
>>>>>>>> (gdb) c
>>>>>>>> Continuing.
>>>>>>>> [New Thread 13840.0x31b8]
>>>>>>>> [Switching to Thread 13840.0x3628]
>>>>>>>>
>>>>>>>> Breakpoint 1, arp_table_search (slirp=0x19f7380, ip_addr=0,
>>>>>>>>     out_ethaddr=0x22f642 "\"") at slirp/arp_table.c:75
>>>>>>>> 75      //    assert((ip_addr & htonl(~(0xf << 28))) != 0);
>>>>>>>> (gdb) bt
>>>>>>>> #0  arp_table_search (slirp=0x19f7380, ip_addr=0, out_ethaddr=0x22f642 "\"")
>>>>>>>>     at slirp/arp_table.c:75
>>>>>>>> #1  0x004bafbd in if_encap (slirp=0x19f7488, ifm=0x2255978)
>>>>>>>>     at slirp/slirp.c:709
>>>>>>>> #2  0x004b8a73 in if_start (slirp=0x19f7380) at slirp/if.c:210
>>>>>>>> #3  0x004b9c9e in ip_output (so=0x2255978, m0=0x0) at slirp/ip_output.c:84
>>>>>>>> #4  0x004bf737 in tcp_output (tp=0x1cac848) at slirp/tcp_output.c:456
>>>>>>>> #5  0x004c09ad in tcp_drop (tp=0x1cac848, err=0) at slirp/tcp_subr.c:225
>>>>>>>> #6  0x004c1182 in tcp_timers (timer=<optimized out>, tp=<optimized out>)
>>>>>>>>     at slirp/tcp_timer.c:287
>>>>>>>> #7  tcp_slowtimo (slirp=0x0) at slirp/tcp_timer.c:88
>>>>>>>> #8  0x004bb6f1 in slirp_select_poll (readfds=0x22fae0, writefds=0x22f9dc,
>>>>>>>>     xfds=0x22f8d8, select_error=2291816) at slirp/slirp.c:433
>>>>>>>> #9  0x0048fb87 in main_loop_wait (nonblocking=0)
>>>>>>>>     at C:/msys/home/User/qemu/vl.c:1436
>>>>>>>> #10 0x00490d10 in main_loop () at C:/msys/home/User/qemu/vl.c:1466
>>>>>>>> #11 qemu_main (argc=0, argv=0x19f5100, envp=0x0)
>>>>>>>>     at C:/msys/home/User/qemu/vl.c:3453
>>>>>>>> #12 0x0049322d in SDL_main (argc=17, argv=0x19f5100)
>>>>>>>>     at C:/msys/home/User/qemu/vl.c:102
>>>>>>>> #13 0x005eb784 in console_main ()
>>>>>>>> #14 0x005eb844 in WinMain@16 ()
>>>>>>>> #15 0x005eb068 in main ()
>>>>>>>> (gdb) frame 4
>>>>>>>> #4  0x004bf737 in tcp_output (tp=0x1cac848) at slirp/tcp_output.c:456
>>>>>>>> 456             error = ip_output(so, m);
>>>>>>>> (gdb) print *tp
>>>>>>>> $1 = {seg_next = 0x1cac848, seg_prev = 0x1cac848, t_state = 0, t_timer = {0,
>>>>>>>>     0, 0, 0}, t_rxtshift = 0, t_rxtcur = 12, t_dupacks = 0, t_maxseg = 1460,
>>>>>>>>   t_force = 0 '\000', t_flags = 0, t_template = {ti_i = {ih_mbuf = {
>>>>>>>>         mptr = 0x0, dummy = 0}, ih_x1 = 0 '\000', ih_pr = 0 '\000',
>>>>>>>>       ih_len = 0, ih_src = {S_un = {S_un_b = {s_b1 = 0 '\000',
>>>>>>>>             s_b2 = 0 '\000', s_b3 = 0 '\000', s_b4 = 0 '\000'}, S_un_w = {
>>>>>>>>             s_w1 = 0, s_w2 = 0}, S_addr = 0}}, ih_dst = {S_un = {S_un_b = {
>>>>>>>>             s_b1 = 0 '\000', s_b2 = 0 '\000', s_b3 = 0 '\000',
>>>>>>>>             s_b4 = 0 '\000'}, S_un_w = {s_w1 = 0, s_w2 = 0}, S_addr = 0}}},
>>>>>>>
>>>>>>> That confirms my theory: the template is not yet initialized.
>>>>>>>
>>>>>>> A shot from the hips: does this patch help?
>>>>>>>
>>>>>>
>>>>>> Yeah the assertion doesn't fail anymore. Thanks.
>>>>>
>>>>> Now I just need to invent some good "why this is correct"... ;)
>>>>>
>>>>> Thanks for testing,
>>>>
>>>> I just have time now to make some tests about -net user.
>>>> And I found that the User mode networking doesn't work anymore after
>>>> your slirp patch series.
>>>
>>> Can you be more verbose? What precisely does not work? Same setup as
>>> before (host, guest, command line, steps to reproduce)? I'm not facing
>>> problems here ATM.
>>>
>>
>> QEMU Revision: 2011-10-12 (latest git give me an Assertion failed:
>> alarm_has_dynticks(t), file qemu-timer.c, line 139. Since there is no
>> change in slirp directory, it should apply to latest git)
>> gcc version:$ gcc -v
>> Target: mingw32
>> Configured with: ../gcc-4.3.3/configure --prefix=/mingw
>> --build=mingw32 --enable-languages=c,ada,c++,fortran,objc,obj-c++
>> --with-bugurl=http://www.tdragon.net/recentgcc/bugs.php --disable-nls
>> --disable-win32-registry --enable-libgomp --disable-werror
>> --enable-threads --disable-symvers
>> --enable-cxx-flags='-fno-function-sections -fno-data-sections'
>> --enable-fully-dynamic-string --enable-version-specific-runtime-libs
>> --enable-sjlj-exceptions --with-pkgversion='4.3.3-tdm-1 mingw32'
>> Thread model: win32
>> gcc version 4.3.3 (4.3.3-tdm-1 mingw32)
>> configure commandline: ./configure --target-list="i386-softmmu
>> x86_64-softmmu mips64el-softmmu" --audio-drv-list=sdl
>> --audio-card-list=ac97,sb16,adlib --disable-linux-aio
>> --enable-vnc-thread --disable-vnc-jpeg --extra-ldflags=-s
>> QEMU Host: Windows XP SP3
>> QEMU Guest: Windows XP SP3, Linux 2.6.38 (NIC: ne2k-pci, e1000)
>> commandline: qemu-system-i386 -hda xp.vmdk -soundhw sb16 -m 320
>> -localtime -usb -usbdevice tablet -net user -net nic,model=ne2k_pci
>>
>> Symptoms:
>> Guest can lease IP from QEMU Virtual DHCP Server, but the outgoing
>> traffic are not working:
>> - 10.0.2.3 Virtual DNS Server doesn't respond any DNS query ("nslookup
>> www.google.com" fails with timeout message, "nslookup www.google.com
>> <host's dns server ip>" and "nslookup www.google.com 8.8.8.8" are also
>> failed)
>> - 10.0.2.2 Virtual Gateway is also unable to connect, only reply
>> pings. Other connections are timeout.
>> - 10.0.2.4 Virtual Samba Server is also unconnectable
>>
>> My last worked build is 2011-9-09 build and it starts to fail since
>> 2011-9-10 build.
> 
> And I'm tried on trying to build every revision. I can find the range
> df00bed0fa30a6f5712456e7add783e470c534c9(ok) to
> 31ff5cc31b10a4a48f854d8e06090c49f92a6720(fail) only.

The second hash appears to be a blob rather than a commit.

I've just re-tested current git head () with a WinXP guest, and it
happily receives dns replies, can ping to the outer world, and connect
to the web.

Could you collect a trace of the host-guest communication, e.g. via the
dump plugin ('-net dump')?

Thanks,
Jan

-- 
Siemens AG, Corporate Technology, CT T DE IT 1
Corporate Competence Center Embedded Linux