[Goodies-dev] [patch] SIGSEGV bug in xfce4-netload-plugin

[Mike Massonnet]

Actually I'm doing some changes on the format_byte_humanreadble
function to display bits now.

(I don't follow the m-l much at all (ccing goodies-dev at xfce.org), and
I check bugzilla occasionnally, when someone bugs me on IRC or by
email usually.)
--
Mike


2014-11-15 23:21 GMT+01:00 J.R. Oldroyd <jr at opal.com>:
> Oh, this is discussed on the xfce mailing lists and bugzilla?
> Why didn't I think of that?!!
>
> Thanks for adding the segfault fix so quickly.
>
> I may have some time next week.  If I do, I'll do the code for
> the bits/bytes and K/Ki changes and send you another patch.  But
> don't go holding your breath!
>
>         -jr
>
> On Sat, 15 Nov 2014 23:16:13 +0100 Mike Massonnet <mmassonnet at gmail.com> wrote:
>>
>> Hi,
>>
>> I just released 1.2.3 with a fix for the segfault :)
>>
>> Regarding the bit/bytes values, there is a recent feature request in bugzilla:
>> https://bugzilla.xfce.org/show_bug.cgi?id=9932
>>
>> Regarding the 1000/1024 values, there is alos one:
>> https://bugzilla.xfce.org/show_bug.cgi?id=9932
>>
>> Just don't have time to hack that into the plugin, but there could
>> definitely be options in the settings dialog.
>>
>> Cheers,
>> --
>> Mike
>>
>>
>> 2014-11-15 21:38 GMT+01:00 J.R. Oldroyd <jr at opal.com>:
>> > Cool, thanks.
>> >
>> > After sending that patch, I also thought that this tool should
>> > really display the numbers in B, KB, MB, etc, rather than B, KiB,
>> > MiB, etc.
>> >
>> > While storage is usually described in 1,024 byte blocks, moving
>> > data is usually described in 1,000 block sizes.  Ethernet is
>> > 100 Mbps, 1 Gbps, 10Gbps etc in 1,000 multiples; WiFi is 54, 150
>> > and 300 Mbps; data lines like T1s, T3s are all in multiples of a
>> > basic DS0 which is 64,000 bps.  Seems odd that this tool uses a
>> > different multiple.
>> >
>> > Of course, folks may have their personal preference.  Perhaps we
>> > should have options:
>> >         Display unit:           bits -or- bytes
>> >         Display multiple:       K (1,000) -or- Ki (1,024)
>> >
>> > Cheers,
>> >         -jr
>> >
>> >
>> > On Sat, 15 Nov 2014 21:20:14 +0100 Mike Massonnet <mmassonnet at gmail.com> wrote:
>> >>
>> >> Hello,
>> >>
>> >> I'm looking into it, thanks for the report.
>> >>
>> >> --
>> >> Mike
>> >>
>> >>
>> >> 2014-11-14 16:28 GMT+01:00 J.R. Oldroyd <jr at opal.com>:
>> >> > Hi,
>> >> >
>> >> > Not sure who is the developer for this at the moment, so sorry if
>> >> > it isn't you.  Maybe you could forward this?  Also including the
>> >> > FreeBSD port maintainer.
>> >> >
>> >> > Since doing a major upgrade here some weeks back (FreeBSD 10.0,
>> >> > xfce-4.10, netload-plugin-1.20), I've been having regular (as in 10x
>> >> > per hour) crashes of xfce4-netload-plugin.  The symptom is a pop-up
>> >> > saying that the plugin "left the panel unexpectedly" and do I want to
>> >> > restart it.  Very annoying to have that all the time.
>> >> >
>> >> > The log revealed:
>> >> >
>> >> > Nov 13 11:46:56 shibato kernel: pid 12782 (xfce4-netload-plugi), uid 1001: exited on signal 10 (core dumped)
>> >> >
>> >> > and, after updating to the latest plugin (1.22), today:
>> >> >
>> >> > Nov 14 08:24:38 shibato kernel: pid 28201 (wrapper), uid 1001: exited on signal 11 (core dumped)
>> >> >
>> >> > And gdb:
>> >> >
>> >> > $ gdb /usr/local/lib/xfce4/panel/wrapper wrapper.core
>> >> > ...
>> >> > Core was generated by `wrapper'.
>> >> > Program terminated with signal 11, Segmentation fault.
>> >> > ...
>> >> > (gdb) bt
>> >> > #0  0x00000008026d5e85 in strlcat () from /lib/libc.so.7
>> >> > #1  0x000000080845cfdc in xfce_panel_module_construct () from /usr/local/lib/xfce4/panel/plugins/libnetload.so
>> >> > #2  0x000000080845ba76 in xfce_panel_module_construct () from /usr/local/lib/xfce4/panel/plugins/libnetload.so
>> >> > #3  0x00000008020e3004 in g_list_sort_with_data () from /usr/local/lib/libglib-2.0.so.0
>> >> > #4  0x00000008020e6702 in g_main_context_dispatch () from /usr/local/lib/libglib-2.0.so.0
>> >> > #5  0x00000008020e6aa3 in g_main_context_pending () from /usr/local/lib/libglib-2.0.so.0
>> >> > #6  0x00000008020e6dcf in g_main_loop_run () from /usr/local/lib/libglib-2.0.so.0
>> >> > #7  0x0000000800b88a0f in gtk_main () from /usr/local/lib/libgtk-x11-2.0.so.0
>> >> > #8  0x00000000004037df in main ()
>> >> > (gdb)
>> >> >
>> >> > which is not a very helpful backtrace, but grepping shows only one
>> >> > call to strlcat() in this plugin code in utils.c.
>> >> >
>> >> > Examining that code led me to the attached patch.
>> >> >
>> >> > The main error and fix is that the array index, uidx, is compared to
>> >> >     sizeof(unit_names)
>> >> > but it should be
>> >> >     sizeof(unit_names)/sizeof(unit_names[0])
>> >> > I think this limit check is being hit if the plugin doesn't run
>> >> > briefly and during that time there is a burst of network traffic
>> >> > which yields an unusually high number.
>> >> >
>> >> > While here, I also noticed that the initial number is already divided
>> >> > by 1024 and shown as KiB which means network transfers of a few bytes
>> >> > cannot be seen.  I changed it so just a few bytes are shown.
>> >> >
>> >> > And I also added some more units to the array - probably won't be
>> >> > needed any time soon, so keep or leave this part as you wish.
>> >> >
>> >> > I have not seen any further crashes since applying this patch.
>> >> >
>> >> >         -jr
>> >
>